Shifting agricultural practices to produce sustainable, low carbon intensity feedstocks for biofuel production

The two-year rotation of corn and soybean adopted as BAU practice in this analysis results in higher corn and soybean yields compared to the respective monocultures (Behnke et al 2018). The county-level yield information for both crops was collected from USDA NASS (United States Department of Agriculture 2019) and utilized as inputs for SOC modeling with the assumption that their yields were recorded under the corn-soybean rotation. This assumption is justified by the prevalence of corn-soybean rotation in most of US Midwestern states, particularly those that produce large amounts of ethanol (Green et al 2018).

Winter CC planting in a corn-soybean rotation is gaining popularity as a conservation practice that improves SOC stock and provides agronomic and environmental benefits to subsequent cash crops (Marcillo and Miguez 2017). Winter rye (Secale cereale L.) and hairy vetch (Vicia villosa Roth) are considered in this analysis. The latter is a legume crop and can fix N from air into soil and provide an N benefit in the form of reduced N fertilizer requirement. The N benefit from a legume CC can be as high as 45 kg N/ha (United States Department of Agriculture Natural Resources Conservation Service 2014). Qin et al (2015) has implemented winter rye CC into the GREET by compiling the data on energy and material consumption due to winter rye cultivation and county-level rye yields (Feyereisen et al 2013). We collected the same type of information for hairy vetch CC (Undersander et al 1990, PennState Extension 2010).

Tillage practices are classified by the percentage of residue remaining on the soil surface after planting. The rest of the residue is tilled back into the soil, since we assumed no stover harvest from the field. USDA reports the share of corn-planted area and the percentage of residue remaining for four tillage types, namely, conventional tillage (CT), reduced tillage (RT), mulch tillage (MT), and no tillage (NT).

To reduce the complexity of CI calculation and SOC modeling, we combined the RT and MT categories in the USDA classification into a single RT category (more details are available in the Tillage subsection of SI). Correspondingly, USDA's nation-averaged and state-averaged share of different tillage types can be applied in this analysis (table S5 in SI).

Compiling regionalized inventory on tillage share is necessary because different tillage practices incur different energy use rates. The diesel fuel requirements for various farming operations were obtained (University of Nebraska–Lincoln Institute of Agriculture and Natural Resources 2019) and the average diesel use rate was calculated for each tillage type (i.e. CT, RT, NT) (table S6 in SI). Note that the energy use rate for CT is almost 3.5 times as high as that for NT practice, providing the justification for incorporating this variation.

N fertilizer management practices are crucial to improving crop yields and reducing N losses. While optimal rate, type, timing, and placement of N fertilizer are important factors, efforts have been made recently to develop stable EEF. It is reported that from 2005 to 2010, the use of EEF increased from 8.5% to 12.5% (Baranski et al 2018).

Here, we have evaluated one type of EEF—nitrification inhibitor that slows down the nitrification process where fertilizers are broken down to produce nitrates and N₂O (ICF International 2016). According to a meta-analysis (Thapa et al 2016), nitrification inhibitor reduced N₂O emissions compared to conventional N fertilizer by 30% and increased crop yields by 7%. These values are leveraged when compiling the regionalized inventory and modeling SOC change, by adjusting the crop yields and N₂O EFs. Nevertheless, GHG emissions from the production and transportation of nitrification inhibitor are excluded from this analysis, since their contributions to the cradle-to-farm-gate emissions are minor (ICF International 2016).

Animal manure can be used as an organic fertilizer to improve soil quality by adding organic carbon and nutrients (e.g. N and phosphorus). County-level manure application rates and types were used for SOC modeling (Xia et al 2019). Manure application has already been implemented into the GREET model by compiling the data on energy consumption during manure transportation and application. The transportation distance for manure was estimated to be 0.367 mile and the transportation energy intensity was 10,416 Btu/ton manure/mile (Qin et al 2015). In term of application, it is assumed that 73.7% of the manure is applied via spreading and the rest is applied through direct injection (Energy Systems. Argonne National Laboratory 2018). Besides, the information on nutrient contents in different manure types is also collected.

Manure N has a higher N₂O EF (1.425%) (De Klein et al 2006) as compared to fertilizer N (1.325%) (Xu et al 2019). The trade-off between SOC accumulation and N₂O emissions for manure application is captured in our analysis. On the other hand, the nutrients in manure may reduce the inorganic N and phosphorus fertilizer input, but this possibility has not been considered.

Improving productivity and/or input utilization efficiency of feedstocks is another important technology that has the potential to reduce the CI of biofuels. Recently, Paustian et al (2016) evaluated the potential of deep-rooting crop varieties to sequester SOC and reduce N₂O flux; in this work, the CENTURY model was employed to estimate reference SOC stocks and simulate their changes associated with four hypothetically altered root growth scenarios. In the present study, we have incorporated the deep-rooting corn variety into the SOC modeling by adjusting root distributions, such that an additional 20% of corn root biomass in the 0–30 cm soil layer is moved to a deeper layer.

The yield of corn affects the yield of ethanol. If higher per-acre yield is achieved with the same level of per-acre chemical inputs, the CI of a MJ of ethanol produced from the field is lower. We analyzed two yield trend scenarios: constant and increasing yield. The constant-yield scenario assumes yield based on the 10 year average of county-level corn yield records from 2006 to 2015, while the increasing-yield scenario estimates yield using a simple regression equation derived from county-level corn yield records from 1951 to 2015 (table S7 in SI).

There are additional management practices available to improve the feedstock CI but are not considered in the present analysis. More descriptions are provided in SI.

On the basis of national-average inventory data, feedstock production emits 28.5 g of CO₂e per MJ of ethanol produced. Our results using regionalized inventory data compiled for the nine corn-farming states demonstrate a large degree of CI variation (figure 1). Compared to the baseline, Ohio has 12% higher GHG emissions while Minnesota has 13% lower. This is because an average Ohio corn grower uses more farm inputs and energy to produce a bushel of corn; while Minnesota farmers use less (table S2).

Zoom In Zoom Out Reset image size

When compared to the baseline scenario, the yield increase scenario reduces the CI by 22%, as expected because the increase in yield means that fewer inputs are required for each bushel of corn. Similarly, CI reduction is observed when nitrification inhibitor is used, since its application is associated with a 7% yield increase and a 30% reduction in N₂O emissions from N fertilizer (Thapa et al 2016).

If a vetch CC is planted after soybean in addition to the previous corn-rye year, more GHGs are emitted. This is because of the additional energy, fertilizers, and herbicides required for vetch cultivation, even after accounting for the N benefit from a legume CC. Similar trends are observed when comparing the scenarios where only rye CC is used and where manure is applied, since both scenarios require additional application of energy.

Finally, tillage intensity changes the cradle-to-farm-gate emissions. The NT practice consumes less energy and incurs less GHG emission compared to the average tillage practice. Since we assumed that changing to deep-rooting corn would not result in additional farming energy inputs, its GHG emissions remain the same as with the baseline scenario. It is important to reiterate that the discussion here (section 3.1.1) deals with farming energy and material inputs only. The net cradle-to-farm-gate GHG emission will be discussed separately in section 3.1.3.

Using the parameterized CENTURY model (section 2.1), the effects of farming practices/technologies on SOC changes can be quantified. Figure 1 indicates that diverse land management practices can change the SOC stock. Practices and scenarios that lead to an increase in corn yield (i.e. yield increase scenario and nitrification inhibitor application scenario) contribute positively to SOC sequestration (i.e. negatively to SOC emissions). This finding is reasonable, since the SOC results are represented as relative changes compared to the baseline scenario, where corn yield is assumed to be constant.

The implementation of CC(s) contributes positively to SOC stocks. The corn/rye-soybean system can sequester, on average, 9 g of CO₂ per MJ of ethanol produced. If a corn/rye-soybean/vetch rotation is used, an additional 17.6 g of CO₂ can be sequestered, clearly showing the positive effect of CC on SOC preservation.

In terms of the tillage type, NT practice contributes positively to SOC preservation, since it disturbs the soil less, leading to a net CO₂ sequestration of 4.6 g compared to the baseline value. The application of manure can also increase the SOC storage by 11.6 g per MJ of ethanol produced from corn harvested from that field. Deep-rooting corn adds more root biomass to soil, thus also helping to improve SOC content. On the basis of these results, the best practice for SOC conservation is no-till, deep-rooting, corn/rye-soybean/vetch rotation, with manure and nitrification inhibitor application, under a yield increase scenario.

Figure 2 shows large variations in the relative SOC sequestration achieved across Midwestern counties by shifting from baseline to best practice in terms of SOC conservation. The largest within-state variation was found in Nebraska, where the difference is 4453 kg CO₂ per hectare per year. Even for the state that has the smallest within-state variation, i.e. Minnesota, the difference can still be as large as 2058 kg CO₂ per hectare per year. By converting the units and distributing this impact to the resulting fuel, we calculate that changing practice results in a 43 g difference in CO₂ emission per MJ of ethanol produced using the national-average corn yield of 166 bushels per acre. These results indicate the necessity of conducting localized assessment for SOC changes, since they are greatly impacted by local environmental parameters.

Zoom In Zoom Out Reset image size

While sophisticated process-based models are applicable to provide SOC quantification with less cost, the premise behind this assertion is that these models can be continuously calibrated and validated with databases of regional variations in management practices and their impacts on SOC changes for biofuel feedstock production. It should also be noted that SOC modeling results are simulated under certain assumptions (e.g. a 30 year-averaged climatic condition) and scenarios (e.g. constant yield versus increasing yield) related to model inputs, and thus should be interpreted with caution. Although longer simulation period or changing future climate may lead to different results, this 30 year simulation under current climate conditions provides sufficient justification for further evaluation through agronomic and policy experiments. Alternative climate scenarios require additional measurements and are beyond the scope of this simulation.

Alternatively, meta-analytic approaches similar to the IPCC Tier 1 methodology (De Klein et al 2006) can help to address the magnitude and uncertainty of SOC changes in agriculture to some degree (Paustian et al 2016). Eventually, field-measured SOC data may be used for CI certification.

The cradle-to-farm-gate GHG emissions are calculated by combining those from LMC-induced SOC change and those due to energy and material consumption. The combination of these two sets yields a comprehensive assessment of biofuel GHG emissions.

For the scenario involving both rye and vetch CC, the CI of feedstock production could be as low as 7.5 g CO₂e/MJ, a 74% reduction compared to the baseline scenario. This finding again confirms that LMC has a large impact on the overall cradle-to-farm-gate GHG emissions.

Figure 1 also demonstrates the comparative magnitude of input-induced variation versus management-induced variation. By comparing the regional variations (i.e. the height of the error bar) to the national-average value for each scenario, we can infer that variations due to regionalized inputs and spatially explicit SOC modeling can be at least as great as the national-average values. This indicates the need for a land management practice database with high spatial resolution in order to support CI certification at the field level.