Biomass yield potential on U.S. marginal land and its contribution to reach net‐zero emission

Bioenergy with carbon capture and geological storage (BECCS) is considered one of the top options for both offsetting CO2 emissions and removing atmospheric CO2. BECCS requires using limited land resources efficiently while ensuring minimal adverse impacts on the delicate food‐energy‐water nexus. Perennial C4 biomass crops are productive on marginal land under low‐input conditions avoiding conflict with food and feed crops. The eastern half of the contiguous U.S. contains a large amount of marginal land, which is not economically viable for food production and liable to wind and water erosion under annual cultivation. However, this land is suitable for geological CO2 storage and perennial crop growth. Given the climate variation across the region, three perennials are major contenders for planting. The yield potential and stability of Miscanthus, switchgrass, and energycane across the region were compared to select which would perform best under the recent (2000–2014) and future (2036–2050) climates. Miscanthus performed best in the Midwest, switchgrass in the Northeast and energycane in the Southeast. On average, Miscanthus yield decreased from present 19.1 t/ha to future 16.8 t/ha; switchgrass yield from 3.5 to 2.4 t/ha; and energycane yield increased from 14 to 15 t/ha. Future yield stability decreased in the region with higher predicted drought stress. Combined, these crops could produce 0.6–0.62 billion tonnes biomass per year for the present and future. Using the biomass for power generation with CCS would capture 703–726 million tonnes of atmospheric CO2 per year, which would offset about 11% of current total U.S. emission. Further, this biomass approximates the net primary CO2 productivity of two times the current baseline productivity of existing vegetation, suggesting a huge potential for BECCS. Beyond BECCS, C4 perennial grasses could also increase soil carbon and provide biomass for emerging industries developing replacements for non‐renewable products including plastics and building materials.


| INTRODUCTION
Currently, terrestrial plants are estimated to assimilate 113 Pg C year −1 through photosynthesis, about 10 times more than the estimated 11 Pg C year −1 released by humans through burning fossil fuels and net land-use change (Masson-Delmotte et al., 2021).This highlights the potential of vegetation as a powerful means to remove CO 2 from the atmosphere.This value can only be realized if the assimilated carbon is stored in some way or used to replace fossil carbon uses.Reforestation and afforestation provide one means for storing CO 2 longterm in the form of woody trees (Shukla et al., 2022).Another emerging opportunity is the use of highly productive and sustainable C4 perennial grasses, which provide an annual crop of biomass (Dohleman et al., 2012;Heaton et al., 2008).Their perennial root and rhizome systems bind the soil, preventing erosion while adding very substantial amounts of carbon to the soil (Clifton-Brown et al., 2007;Zhao et al., 2023).These perennial grasses were originally developed as a source for renewable biofuels to replace fossil fuels.Increasingly they are finding markets in replacing geologic oil derived plastics and building materials (Liu et al., 2017).The greatest value in offsetting CO 2 emissions would be obtained by not only replacing fossil fuels, but also by capturing and storing the CO 2 emitted in their combustion (Rosa et al., 2021;Shukla et al., 2022).Land and techno-economic analyses suggest that removal of 1.6 Pg C year −1 , or about 15% of current emissions, could be attained with this use of biomass by 2050 at acceptable costs (Shukla et al., 2022).Such bioenergy with carbon capture and storage (BECCS) has been highlighted in the most recent report of the Intergovernmental Panel on Climate Change as one of the top opportunities for CO 2 removal from the atmosphere (Shukla et al., 2022).Additionally, bioenergy and carbon sequestration were key elements of the most recent U.S. Farm Bill 2018 (Agriculture Improvement Act of 2018).In this context, these highly productive C4 perennials may have particular value for the mitigation of CO 2 emissions in the U.S.
Despite the large area of Midwest row-cropland, the eastern half of the Contiguous United States (CONUS) also contains a large amount of marginal land unsuitable for row crop production due to its erodibility or poor soil quality (Cai et al., 2011;Jiang et al., 2021).Most economically marginal land in the 48 contiguous states, i.e., land that has low profitability and is currently not in food crop production, is suitable for growing energy crops without irrigation is east of 105° W longitude (Jiang et al., 2021).This area includes land that was abandoned from row crop use following the dust bowl era and other land used for low-intensity grazing (Cook et al., 2009).Importantly, much of this growing area coincides with regions suited to geological sequestration of CO 2 (Beerling et al., 2018;U.S. Geological Survey, 2013).A further benefit shown by biophysical analysis of land use change, is that planting these perennials could cause a summer cooling of the region of up to 1°C, while causing an increase in precipitation and decrease in water vapor pressure deficit, all serving to offset predicted climate change (He et al., 2022).
Here, we show that given the exceptional scale of this marginal land area, we can re-imagine its use for capturing a large amount of atmospheric CO 2 (Valentine et al., 2012).Such large-scale land use change for BECCS has been shown to be economically viable in contributing to emission reduction below 2 or 1.5°C warming scenarios (Fajardy et al., 2021).Early examples of the production and utilization of BECCS are in place (Jaiswal et al., 2017), but further adoption of this land-use change strategy will depend on its economic viability in terms of projected yields and the year-toyear stabilities of those yields in relation to the price per ton placed on carbon sequestration (Clifton-Brown et al., 2023;Zhang et al., 2023).Since the rollout of such land use change and development of markets will take time at scale, it is important to account for regional climate change that could occur over the next 20 years in projecting which bioenergy crops would be best where.
Given the large climatic and soil variation in the eastern half of CONUS, different crops suitable for local climate and soil conditions will need to be selected to maximize biomass this biomass approximates the net primary CO 2 productivity of two times the current baseline productivity of existing vegetation, suggesting a huge potential for BECCS.Beyond BECCS, C4 perennial grasses could also increase soil carbon and provide biomass for emerging industries developing replacements for non-renewable products including plastics and building materials.
bioenergy crops, biomass production, C4 crop growth modeling, carbon capture and storage, CO 2 emission reduction, marginal land use and economic viability.The focus of this work is on the three C4 perennial grass crops that have emerged as highly productive: (1) Miscanthus, a relative of sugarcane from East Asia but more suited to colder climates, with a natural distribution extending into Siberia; (2) Energycane, a low sugar and high fiber derivative of sugarcane that is less environmentally demanding than its antecedent; and (3) Switchgrass, a native perennial grass which has received the most attention in the U.S. and includes a wide range of cultivars.
To understand yield potential and stability, and in turn sequestration potential for CO 2 under current and future climate conditions, we extended and validated a semi-mechanistic crop model originally developed for Miscanthus to all three crops.Using one modeling framework avoids the risk of compounding model differences with species differences.The climate data for driving the crop model were generated using a regional climate model to simulate 15 years for both the recent (2000)(2001)(2002)(2003)(2004)(2005)(2006)(2007)(2008)(2009)(2010)(2011)(2012)(2013)(2014) and the future (2036-2050) periods under the climate scenario SSP5-8.5 (Masson-Delmotte et al., 2021).With this recent and future climate data, soil type data from the United States Geological Survey, and the BioCro crop modeling framework parameterized, calibrated, and validated for the three crops, we predicted the yields and yield stabilities under current and projected future conditions.Further, we combined marginal land use data with the projected biomass yield to quantify the amount of biomass that can be produced without competing with land used for food and feed crops.

| Model description and calibration
The model BioCro (Lochocki et al., 2022) was used in this study to simulate the growth of three biomass crops: Miscanthus, Switchgrass, and Energycane.This version of BioCro incorporates the Collatz et al. (1992) to predict net C4 photosynthesis and stomatal conductance, as in our earlier implementations of BioCro in C4 crops (Jaiswal et al., 2017;Miguez et al., 2012).This is incorporated into a 10-layer canopy model that dynamically categorizes the proportions of sunlit and shaded leaves in each layer for light interception and assimilation at an hourly time step accounting for the latitude and solar position.Net assimilation is partitioned between plant organs with a scheme based on logistic functions.The growth and senescence of leaf, stem, rhizome and root biomasses were so predicted throughout the growing season.
The parameterization of the model was previously implemented for Miscanthus and switchgrass (He et al., 2022;Miguez et al., 2012).In this study, we re-parameterized the two crops with an updated version of BioCro (Lochocki et al., 2022) and extended the parameterization to energycane.The partitioning coefficients of all three crops were calibrated against observed data (Table S1).After the single-site parameterization, the model was applied to multiple locations where observed data were available.The model performance was evaluated using root mean square error (RMSE) and the Pearson's correlation coefficient.The key model parameters for each crop are summarized in Table S1.The calibration and validation data are available in the GitHub repository (https:// github.com/ Matth ews-Resea rch-Group/ bioma ss-crops -margi nal-land).

| Regional climate data and future scenario
A simulation period of 15 years was selected to represent the recent (2000)(2001)(2002)(2003)(2004)(2005)(2006)(2007)(2008)(2009)(2010)(2011)(2012)(2013)(2014) and future scenarios (2036-2050), respectively.We used a recently developed climate scenario SSP5-8.5 that represents the high end of the range of future emission pathways, corresponding to RCP8.5.Despite being at the high end it continues to be the best match scenario to emissions (Schwalm et al., 2020).The global climate simulations of the Max Planck Institute for Meteorology Earth System Model (Mauritsen et al., 2019) were used to drive a regional climate model CWRF (Liang et al., 2012) at a spatial resolution of 30 × 30 km 2 .Further details on the CWRF model can be found in our previous work on Miscanthus (He et al., 2022).CWRF's threehourly climate data were interpolated to the hourly time step and then used to drive the BioCro model to predict yields of the three crops in CONUS.

| Growing period and harvest
The starting and ending day of the year for growing seasons were estimated using CWRF's 2-m temperature records to determine the first and last days of a given year when a frost condition occurred (i.e., temperature below 0°C).This was done for each year during the past and future scenarios, resulting in different lengths of growing season for each location and each year.This is a widely used but simplified method on estimating the growing season period.A single frost event may not be sufficient to terminate growth completely and several consecutive frost days may be a better model assumption (Magenau et al., 2022).However, more data are required to quantify this mechanism for multiple locations in the US.
Once senesced at the end of growing season, Miscanthus stands will progressively lose dead biomass through fragmentation.An average loss of 0.07 t/ha day −1 (Heaton, 2006) was used for Miscanthus and March 1 was assumed to be the harvest date in our simulations to allow nutrient recycling during winter.Therefore, for each location, the number of days for senescence was calculated from the last day of the growing season to the harvest date.A grid point is discarded if the harvestable biomass became negative.For switchgrass and energycane, the harvest dates were assumed to be the last day of the growing season without experiencing winter loss, consistent with existing agronomic practice.

| Recent and future plant hardiness zone
We used a temperature-based criterion called the Plant Hardiness Zone Map (https:// plant hardi ness.ars.usda.gov/ ) to determine the suitable area for Miscanthus and energycane.In general, the widely used Miscanthus × giganteus Illinois clone on which our parameterization is based can be grown in zones 5-8 (Dong et al., 2019) and energycane can be grown in zones 8-11 (Viator et al., 2010).This choice was based on an assumption that in warm southern areas, sustainable cultivation of Miscanthus would not be possible since this is an inadequate signal for remobilization of biomass reserves to the rhizome for regrowth of the next year, while energycane is chilling sensitive (Burner et al., 2015).Since there are uncertainties in determining the boundary zones for Miscanthus, we used the hardiness zones 5a-8a as the suitable region for Miscanthus (Figure S1).In this case, any areas south of the 8a were suitable for enerycane but not Miscanthus.Switchgrass was assumed to be able to grow across the entire CONUS, given the availability of cultivars adapted to a wide range of climates (LeBauer et al., 2018).The inter-annual averaged minimum temperature was calculated using the CWRF's temperature data for the recent and future scenarios, respectively.

| Optimal crop allocation
The hardiness zone above sets an initial constrain on where each crop can be grown.Since switchgrass was allowed to grow in all places, there were grid points that could be assigned to any one of the three crops based on their productivity.We thus compared the interannual average yield among the three crops and allocated each grid point to the crop with the highest yield.This allocation was done for both recent and future climates.

| Marginal land use
We used recently developed marginal land use data that has a total marginal land area of 58.3 million ha in the CONUS (Jiang et al., 2021).This includes cropland in transition with confidence and with uncertainty.The cropland in transition with confidence consists of fallowed cropland, expanded cropland, and abandoned cropland as defined by Jiang et al. (2021).The cropland in transition with uncertainty may not be economically marginal but has similar biophysical and environmental characteristics as the ones assigned with confidence.About 44 million hectares are in the eastern region of CONUS as defined in this study (Figure S2).The original data were at a 30-m resolution and was aggregated into the 30-km resolution to match the climate data used in this study (Figure S2).The total biomass across the entire marginal land was calculated by the summation of the product of yield (t/ha) and marginal land area (ha) per grid point.

| Net primary productivity calculation
We used the historical net primary productivity (NPP) of U.S. (Foley et al., 1996) and masked it with the marginal land of eastern U.S. (Figure S2).Then the baseline NPP was calculated by an average of all grid points on the marginal land.For the model result of each bioenergy crop, the potential NPP were estimated by averaging the peak biomass of each year across the grid points on marginal land.The peak biomass included four organs: stem, leaf, root and rhizome.

| Biomass energy efficiency for power generation and the CO 2 offset
We used a biomass energy content of 17 MJ/kg and a 30% conversion efficiency to electricity (Evans et al., 2010).A previous lifecycle analysis estimated the greenhouse gas emissions for electricity generation at about an average of −827 gCO 2 eq/kWh if using biopower with CCS (Sathaye et al., 2011).The total atmospheric CO 2 offset was then calculated by converting the predicted total biomass produced across the U.S. marginal land to the CO 2 equivalent.

| Biomass to ethanol
We used a conversion efficiency of 282.2 L/Mg biomass of 15% moisture content to calculate the cellulosic ethanol yield from Miscanthus (Dwivedi et al., 2015).The same conversion rate was assumed for switchgrass.A conversion rate of dry biomass of 13% was assumed for energycane (Kim & Day, 2011).

| Optimized crop allocation, yield and yield stability
The BioCro model performance versus measured yields on multiple sites for Miscanthus, switchgrass, and energycane (Figures S3-S5) have RMSE of 8.0, 5.8 and 14.0 t/ ha and correlation coefficients of 0.47, 0.33 and 0.54, respectively.The observed biomass measurements from these multiple field sites include a wide range of cultivars, fertilization strategies, pest control and other management practices.The resulting variability caused by these practices are not accounted for in the BioCro model.
When comparing the predicted inter-annual average biomass yields (Figure 1), we found that Miscanthus dominated the Midwest, switchgrass occupied the northern area, and energycane dominated the southern part of this region in both the recent and future time periods.The average predicted annual yield of Miscanthus, switchgrass, and energycane were 19.1 ± 3.4, 3.5 ± 1.0, and 14.0 ± 3.7 t/ ha across the grid points assigned to them for the recent period.Future warming pushed the boundaries of the hardiness zones that were suitable for growing Miscanthus and energycane further north, increasing both the areas for growing Miscanthus and energycane but decreasing the amount of land optimal for switchgrass.Miscanthus yield in the future was predicted to remain unchanged in most of the Mid-west and mid-east regions.In the western part of the studied area, although the radiation increased in spring and summer (Figure S6), the yield decreased in both Miscanthus (16.8 ± 3.0 t/ha) and switchgrass (2.4 ± 0.8 t/ha), mainly due to drier soil and higher water stress impacts on photosynthesis and leaf growth (Figure S7).In comparison, energycane yield (14.7 ± 3.8 t/ ha) increased as water stress was reduced in the Southeast region (Figure S7).Significant yield changes were seen at the northern and southern hardiness boundaries, where yield increased in the north when switchgrass was replaced by Miscanthus and decreased in the south when Miscanthus was replaced by energycane (Figure 1c).
In both the recent and future scenarios, the predicted Midwest yields were more stable (Figure 2) due to abundant rainfall and higher clay content in the soil facilitating better water holding capacity (Figures S8 and S9).In the future scenario, annual yield became less stable in the western part and more stable in the Midwest and Southeast (Figure 2c).At some locations, the changes were greater due to the change of crop type.For example, at the boundary strip separating Miscanthus and energycane, there was a large difference in the yield stability (Figure 2c) as the energycane yield in the future was predicted to be less stable than the Miscanthus yield in the past.

| Biomass production on the marginal land
Over the entire marginal land in the eastern part of CONUS (Figures S2 and S10), the total biomass production of Miscanthus was estimated to be 0.39 ± 0.02 billion tonnes per year for the recent period with a land of 19.5 million ha and 0.36 ± 0.01 billion tonnes per year for the future with a land area of 20.9 million ha (Figure 3).Switchgrass produced 0.09 ± 0.01 billion tonnes per year on 16.8 million ha for the recent and 0.04 ± 0.01 billion tonnes per year on 11.4 million ha for the future (Figure 3).Energycane produced 0.14 ± 0.01 billion tonnes per year on 7.4 million ha for the recent and 0.2 ± 0.02 billion tonnes per year on 11.4 million ha for the future (Figure 3).
In the future scenario, the total biomass of Miscanthus was reduced by 8% and its land area was reduced by 7%.Switchgrass biomass was more than halved and its land area was reduced by 32%.In comparison, the total biomass of energycane was increased by 42% and its land area was increased by 24%.enormous potential of this land for carbon from the atmosphere if used for BECCS, given that much of this land coincides with geology suitable to geologic carbon storage (U.S. Geological Survey, 2013).CO 2 reductions would also be achieved if this biomass fueled emerging industries developing replacements for non-renewable resources in addition to the production of biofuels.
BECCS has been highlighted as one of the top potential contributors to achieving net-zero emissions (Shukla et al., 2022).BECCS provides heat and energy that might otherwise be provided by fossil fuels, while capturing the CO 2 from combustion of the biomass for geological sequestration.At the same time the highly productive perennial crops, as examined here, on marginal land would add more carbon to the soil and through changes in evapotranspiration and albedo are predicted, at scale, to cause regional cooling, precipitation increase and increase in humidity, which would benefit food and feed crops in the region (He et al., 2022).However, deployment requires careful consideration of crop choice and location to make efficient use of limited land resources to minimize adverse impact on the delicate food-energy-water nexus and biodiversity (Davis et al., 2018;Field et al., 2008).We showed that three perennial biomass crops suited for marginal land across the eastern CONUS can produce about 0.6-0.62 billion tonnes of biomass per year.If using the biomass conversion for power generation with CCS, this would result in 703-726 million tonnes of atmospheric CO 2 captured per year.This great potential of carbon capture would offset about 11% of the total emission in the U.S. and would exceed the total current emission from agriculture (EPA, 2023).In fact, this scenario alone could change the total U.S. agricultural emission from a net annual source of 670 Mt CO 2 eq to a net sink of 45 Mt CO 2 eq.Of course, it is unlikely that all the marginal land defined by Jiang et al. (2021) would be used to grow biomass crops for BECCS.However, others have suggested on other criteria for determining marginal land that suggest much larger areas exist in the eastern half of the CONUS (Cai et al., 2011;Yang et al., 2020).
Large scale deployment of these crops would clearly motivate a commercial focus on improvement and utilization of their germplasm pools.For example, our Miscanthus simulations are based on the widely trialed, used and commercially available Miscanthus × giganteus cv."Illinois" clone, but wild germplasm with more than twice the yield potential of this clone has been identified and could be utilized (Clark et al., 2019).Maize and soybean yields have both risen by about 20% over the past 20 years in the U.S. as a result of intensive breeding, genetic modification and increased agronomic inputs (Egli, 2023).These three perennial biomass crops have only had a tiny fraction of this attention.However, the experience with maize and soybeans suggests that substantial yield gains could be achieved if markets were to expand, further increasing the amounts of CO 2 that could be captured.

| Larger NPP increases CO 2 capture and potentially storage
Using the global NPP reported in a previous study (Foley et al., 1996) as a baseline, the current NPP on the marginal land as defined here was estimated to ca. 531 ± 175 g C m −2 year −1 .After the optimal placement of the three crops, the potential NPP of the peak biomass of Miscanthus, switchgrass and energycane on average on the marginal land were 2114 ± 58, 266 ± 25, and 1077 ± 73 g C m −2 year −1 for the recent scenario from 2000 to 2014, respectively.They were estimated to become 2171 ± 94, 206 ± 23, and 1125 ± 60 g C m −2 year −1 for the future scenario from 2036 to 2050.On average, a combination of the three crops could have about twice the baseline NPP.These numbers could be viewed as a minimum, since considerable effort is now being placed in agronomy and breeding of these biomass crops for higher yields and improved sustainability and environmental tolerance.Although switchgrass was outyielded by either Miscanthus or energycane at most locations, breeding of this crop is currently the most advanced of the three and new cultivars might change this map (Casler, 2020;Lovell et al., 2021).Another major advantage of switchgrass is that it is planted using direct sowing, while Miscanthus and energycane rely mostly on clones.
This large, predicted increase in NPP could contribute to the goal of increasing soil organic carbon as part of the 4 per 1000 initiatives voluntary action plan adopted by the United Nations Framework Convention on Climate Change (UNFCCC) (Soussana et al., 2019).Previous studies including both modeling and field trials have shown that replacing annual crops with Miscanthus or switchgrass could significantly increase soil organic carbon (Hudiburg et al., 2015;Kantola et al., 2017), though there is still more work needed to fully understand this process.While some studies suggest that soil carbon sequestration could be largely limited by plant photosynthetic capacity (Janzen et al., 2022), our study demonstrated that growing perennials with high photosynthetic assimilation could circumvent this limitation and contribute to the 4 per 1000 initiatives globally.
In fact, Miscanthus has been shown to be a carbonnegative bioenergy crop in the U.S. Midwest (Dwivedi et al., 2015), and even without BECCS we expect it to remain carbon-negative across the marginal land considered in this study (Davis et al., 2012).Similar benefits of being carbon-negative have also been found in a wide range of studies in switchgrass (Monti et The impact of converting pastureland to energycane on greenhouse gas balance may be positive or negative (Gomez-Casanovas et al., 2018).However, the positive impact of climate change on energycane cultivation shown in this study, and biotechnological approaches to produce oil from energycane, may make energycane more beneficial for greenhouse gas savings (Kumar et al., 2021).Further, a recent study showed that applying ground basalt to miscanthus fields doubled the carbon negative intensity of Miscanthus over a 5-year period (Kantola et al., 2023).This would further increase net CO 2 removal from the atmosphere, with or without BECCS.This biogeochemical treatment of the land to enhance CO 2 removal would likely apply across all three crops.

| Biofuel potentials and beyond
The ethanol equivalent of the total biomass are about 48 and 44 billion gallons for the recent and future periods respectively.Our estimated volume of ethanol suggests that we could easily meet the original goal of 16 billion gallons ethanol produced from cellulosic biomass in 2022 under the 2007 Renewable Fuel Standard (RFS) without competing for land under food production (Hudiburg et al., 2016).Although recent RFS targets have significantly reduced the needs of cellulosic biofuel due to its high production cost (Aui et al., 2021), as the technology becomes more advanced and cheaper, advanced biofuel production could still be significant to national energy interest for green energy, carbon capture and storage in soils (Minasny et al., 2017).For example, GIS-based optimization for biofuel supply chain has been used to minimize annual biomass-ethanol production costs (Lin et al., 2013).This analysis could now be extended by considering the impact of climate change on predicted yield demonstrated in this study.
Beyond biofuels, the biomass from these crops can also be used for biomaterials such as polylactic acid and biobased polyethylene.An advantage of biomaterials over biofuel is its significant efficiency in reducing the nonrenewable energy use and thus greenhouse gas emissions (Bos et al., 2012).In addition, growing these perennials can be environmentally beneficial.For example, nitrogen losses from perennial grass fields were observed to be significantly lower than row crops in the Midwest, contributing to a better soil, water and air quality (Smith et al., 2013).An earlier modeling study looked at the hydrological impact of large-scale land-use conversion to Miscanthus in the Midwest, finding that this land-use change may lead to depleted soil water (Vanloocke et al., 2010).However, that study does not consider the coupled feedback between the crop and climate.Recent work using a two-way coupled climate-crop model, CWRF-BioCro, showed no significant soil water depletion, but an increase in the summer precipitation in part of the Midwest when growing Miscanthus on US marginal land (He et al., 2022).Further, deployment of these crops on a large scale may help mitigate rising temperatures through crop-climate feedback (He et al., 2022) which, in turn, would reduce the projected loss from higher temperatures in annual crops (Lobell & Asner, 2003).

| More work is needed to transition from potential to practical
Marginal land provides a means to growing perennial bioenergy crops without compromising annul crop production.However, the transition to an economy based on biomass production will not be driven by science and technology alone (Clifton-Brown et al., 2023).Socioeconomic analyses have shown that landowners are reluctant to convert marginal land to energy crops, due to various non-economic and behavioral reasons, including cultural and aesthetic preferences (Khanna et al., 2021).To increase the acceptance of transitioning marginal land to energy crop fields, further research and outreach efforts are needed across many disciplines, including socioeconomics, policy, culture, and education.Long-term policy support through incentives will also be needed to overcome barriers in the transition from potential innovations to practical uses (Clifton-Brown et al., 2023).

ACKNOWLEDGMENTS
The model simulations and analyses were conducted using high-performance computation facility of the BioCluster the Carl R. Woese Institute for Genomic (IGB) and the Illinois Campus Cluster the University of Illinois Urbana-Champaign (UIUC).This work was supported by the U.S. National Science Foundation Innovations at the Nexus of Food, Energy and Water Systems under Grant EAR1639327 and EAR1903249, and the U.S. Department of Agriculture-National Institute of Food and Agriculture under Grant 20206801231674.The views expressed in this document are solely those of the authors and do not necessarily reflect those of the funding agency.

F
I G U R E 1 Inter-annual average biomass yields of the three biomass crops simulated for the (a) recent (2000-2014) and (b) future (2036-2050) scenarios.Blue: Miscanthus.Green: Switchgrass.Magenta: Energycane.(c) The yield difference between the future and recent scenarios.The dashed polygons represent the points where the crop type changes from recent to future scenarios.Map lines delineate study areas and do not necessarily depict accepted national boundaries.
We have provided maps of which perennial C4 biomass crops would provide the most yield on marginal land under rainfed conditions, including how climate change will affect these yields and allocations.The results show F I G U R E 2 Yield stability (%) for the three biomass crops simulated for the (a) recent (2000-2014) and (b) future (2036-2050) scenarios.(c) The stability is calculated by the ratio of standard deviations and their means for each simulated period of 15 years.A higher value represents a lower stability.Map lines delineate study areas and do not necessarily depict accepted national boundaries.F I G U R E 3 Total biomass production on marginal land for the three crops averaged over the recent (2000-2014) and future (2036-2050) periods.The error bars represent the interannual standard deviations of the total biomass across the marginal land.