Introduction
Modern industrial agriculture and food systems have been widely recognized as contributing to global environmental challenges, including climate change, biodiversity loss, pesticide pollution, and disruption of carbon and nitrogen cycles (Pingali, Reference Pingali2012; Muller et al., Reference Muller2017). Agriculture is a major source of anthropogenic greenhouse gas (GHG) emissions, accounting for an estimated 9.3 billion tons of carbon dioxide equivalents globally as of 2018, which constitutes about 17% of all GHG across all sectors (FAO, 2021). Emissions due to crop and livestock activities accounted for 57% of total emissions from agriculture, while land use change, such as the conversion of natural vegetation for agriculture, contributed to the remaining 43% (FAO, 2021). Given the high carbon footprint (CF) of agricultural systems, there is an urgent need to mitigate GHG emissions from agricultural systems to enhance climate change mitigation efforts (Clark et al., Reference Clark2020).
Regenerative organic agriculture has emerged as a promising approach to address these challenges through improved management of agricultural land on a global scale (Giller et al., Reference Giller2021; Rempelos, Kabourakis and Leifert, Reference Rempelos, Kabourakis and Leifert2023). Specifically, organic agriculture has been advocated for its potential to reduce the CF of agricultural activities by decreasing GHG emissions. However, broader environmental impacts of organic agriculture remain under-researched, especially in terms of its role in climate change mitigation (Lee, Choe and Park, Reference Lee, Choe and Park2015; Seufert and Ramankutty, Reference Seufert and Ramankutty2017).
Some research has been conducted comparing the effects of different production systems on GHG emissions for paired organic and conventional systems and generally involves two main types of methods and two types of metrics for measuring the performance of these systems. One method involves modeling studies using life-cycle assessments (LCAs) or CFs that rely on activity data and emission factors, often derived from the Intergovernmental Panel on Climate Change (IPCC) guidelines. Alternatively, GHG emissions can be estimated from field-based measurements of major agricultural GHGs (i.e., CO2, N2O, and CH4) taken from paired organic and conventional systems. Modeling studies have the benefit of being relatively low cost to implement and provide a broad picture of emissions in each system, but their accuracy depends on precise representation of physical processes and underlying assumptions. By comparison, field-based measurements are costly yet afford a more accurate accounting of GHG emissions from agricultural land through direct observation (Goglio et al., Reference Goglio2018).
Results of these studies are reported using one of two metric types: either absolute emissions or emissions intensity. Absolute emissions measure the total, cumulative emissions emitted by agricultural systems over a given period, usually reported per hectare of cultivated land (i.e., GHG emissions per hectare). By contrast, emissions intensity normalizes emissions to the output, such as crop yield (e.g., GHG emissions per kilogram of product). Although both metrics are crucial for understanding agriculture’s impact on GHG emissions, they reflect different aspects of agriculture’s overall CF. There is some debate in the research community regarding which metric is most appropriate for comparing emissions between different systems (Meier et al., Reference Meier2015). On one side, emissions per unit area, herein referred to as area-scaled emissions, are the absolute emissions directly from the land surface, which is the metric that is typically used by international bodies such as the IPCC to produce inventories of total cumulative GHG emissions (Chiriacò, Castaldi and Valentini, Reference Chiriacò, Castaldi and Valentini2022). By contrast, emissions intensity or per product emissions, herein referred to as yield-scaled emissions, are the most commonly used metric in agricultural CFs and LCAs because they capture the effect of crop yield in relation to emissions and convey the efficiency of a system in terms of GHG emissions; measuring yield-scaled emissions allows for easy comparison between disparate products and thus has the benefit of conveying to consumers which products are associated with higher emissions. Since organic farming typically has lower yields than conventional farming, differences in yield-scaled emissions could also have implications for land use and emissions associated with crop production (Seufert and Ramankutty, Reference Seufert and Ramankutty2017).
Both CF studies and in-situ measurements have found that organic generally outperforms conventional farming practices in terms of area-scaled emissions but performs inconsistently when examining yield-scaled emissions (Tuomisto et al., Reference Tuomisto2012; Venkat, Reference Venkat2012; Skinner et al., Reference Skinner2014; Aguilera, Guzmán and Alonso, Reference Aguilera, Guzmán and Alonso2015; Meier et al., Reference Meier2015; Reganold and Wachter, Reference Reganold and Wachter2016). The inconsistency in yield-scaled emissions stems from the lower yields typically observed in organic systems, which can result in higher emissions when calculated per unit of output. For instance, one estimate suggests that the yield gap between organic and conventional systems needs to be less than 17% to maintain yield-scaled GHG emissions on par with those in conventional systems (Skinner et al., Reference Skinner2014). Additionally, differences in crop types can influence these comparisons, since organic fruits and vegetables may perform better in terms of yield-scaled emissions, though these findings vary (Lee, Choe and Park, Reference Lee, Choe and Park2015; Chiriacò, Castaldi and Valentini, Reference Chiriacò, Castaldi and Valentini2022).
However, these results are based on relatively few studies, and additional research using LCA and CF methods, along with in-situ measurements of GHG emissions in paired experiments, is required to resolve existing uncertainty and reveal underlying mechanisms contributing to differences between production systems (Skinner et al., Reference Skinner2019). In order to accurately and fairly compare the CFs of organic and conventional farming, future research ought to compare both metric types (i.e., area- and yield-scaled emissions) since these convey different information about the CF of agricultural systems. Therefore, there is a critical need to continue to assess the performance of organic farming as a potential solution for mitigating climate change using multiple metrics and study types.
The Rodale Institute is a pioneer in organic farming research and has decades of data and experience regarding organic farming practices. Rodale’s Farming Systems Trial (FST) has been comparing outcomes between standard organic and conventional farming practices in a northern temperate region for over 40 years and provides a potentially rich source of data for investigating differences in emissions between arable cropping systems. In this study, data from the FST collected between 2008 and 2020 were used to assess differences in the CFs of organic and conventional agricultural systems using a combination of modeling methods. Specifically, the CFs of two organic cropping systems and one conventional cropping system were evaluated in the FST using both IPCC Tier 2 and Tier 3 modeling approaches (IPCC, 2019). Subsequently, the performance of organic and conventional systems in terms of area-scaled emissions for the entire crop rotation and yield-scaled emissions (per kilogram of maize and per unit digestible energy) were evaluated using the two different modeling approaches. This study also included one secondary analysis that determined the impact of including CO2 emissions from composting manure on the CF. We hypothesize that the organic systems will have lower area-scaled emissions, but yield-scaled emissions will be higher for organic systems compared with conventional systems. We further hypothesize that including emissions from composting manure will increase the CF of organic systems.
Methods
Site description
The Rodale Institute’s FST is located in Kutztown, Pennsylvania, USA (40° 33′ 5′′, −75° 43′ 47′′). The site has a warm-summer humid continental climate according to the Kopper–Geiger classification system with an average annual temperature of 12.4 °C and annual precipitation of 1,105 mm. Soils on the site are moderately well-drained silt loam and classified as Oxyaquic Fragiudalf (Soil Survey Staff, 2019). The FST is a long-term research trial that has been running continuously since 1981 to compare the effects of organic and conventional arable cropping systems on soil health, agronomy, and economics. The FST consists of two organic systems and one conventional system. The organic systems are similar but differ in their sources of fertility—one organic system, referred to as the organic legume (LEG) system is a low-input system that relies on leguminous cover crops and green manures as its primary source of fertility. The organic manure (MNR) system is similar to the LEG system in that it relies on cover crops to meet part of its fertility requirements, but it also receives periodic applications of composted manure to mimic an integrated dairy system typical of the region. The experimental design of the FST follows a randomized complete block layout, with each block divided into three subplots corresponding to the cropping systems. Each system is further subdivided into three entry points that represent crops within the rotation. Initially, all plots were managed with intensive tillage until 2008, when reduced tillage (RT) practices were introduced alongside continued full tillage (FT) to assess the effects of tillage on soil health and carbon dynamics. This update allows for comparison between full and conservation tillage practices within each cropping system. The conventional (CNV) system is a 3-year rotation consisting of maize (Zea mays) and soybeans (Glycine max), with 2 years of maize followed by 1 year of soybean. The LEG system has a 4-year crop rotation consisting of maize, soybeans, and small grains (i.e., wheat [Triticum aestivum] and oats [Avena sativa]), with nitrogen fertility being provided by green manure from red clover (Trifolium pratense) and hairy vetch (Vicia villosa), with cereal rye (Secale cereale) used as a winter cover crop. The crop rotation in the MNR system is longer (8–9 years) and consists of the same crops as the LEG system (i.e., maize, soybeans, and small grains), but also includes maize silage and 2 or 3 years of a perennial mixed hay phase consisting of orchardgrass (Dactylis glomerata) and alfalfa (Medicago sativa) as part of the rotation. Winter cover crops in the MNR system include cereal rye and hairy vetch. Additional background information on experimental design, fertility inputs, and farm practices during this 12-year period in the FST can be found in Pearson, Omondi et al. (Reference Pearson2023a).
Data sources
This study relied on data assembled by Pearsons, Omondi et al. (Reference Pearson2023a) as part of an economic analysis to construct CFs of each of the three FST farming systems. This consisted of data collected from 2008 to 2020 on crop data, inputs, and management activities. Crop data consisted of detailed information on crop rotations and crop yields for each year and treatment plot. Data on inputs consisted of information on the date and quantities of synthetic fertilizer applied, the date and quantities of composted manure applied, and the date and quantities of pesticides applied. Data on management activities consisted of types of equipment used and dates for soil tillage, cultivation, planting, and harvesting events. The raw and processed data are provided in the Supplementary Material as part of this publication.
Multiple model approach to carbon footprint analysis
To assess the CFs of organic and conventional production systems, a dual-modeling approach was employed. This approach aimed to capture variations in CF estimates depending on model selection and to provide a comprehensive analysis of GHG emissions from cradle to farmgate. Prior to modeling, emissions were categorized into five different key sources: (i) CO2 emissions from fertilizer application (i.e., from urea fertilization), (ii) N2O emissions from soils and fertilizers, (iii) CO2 emissions from fossil fuels (e.g., combustion during farm operations), (iv) CO2 emissions from fertilizer production, and (v) CO2 emissions from herbicide production. Notably, neither model simulated methane (CH4) emissions from manure application in the MNR system and there was a lack of reliable emission factors in the literature. CH4 emissions were therefore not included in subsequent analyses and, as a result, emissions from the MNR system may be underestimated in the present study. For each category, emissions were converted to metric tons of CO2-equivalent per year (tons CO2-eq year−1) to compare emissions between different gases.
Two separate modeling approaches were used to determine emissions for the five key categories mentioned above. The first model used for CF estimation is regarded as an IPCC Tier 3 method, which typically uses processed-based models or measurement approaches to estimate emissions (IPCC, 2019). The U.S. Department of Agriculture’s (USDA) COMET-Farm tool relies on the process-based DayCent model and was used to estimate emissions of CO2 from fertilizer and N2O from soils, fertilizer, and composted manure applications. The CO2 from fossil fuels for the COMET-Farm modeling was estimated manually using data from Iowa State University extension on fuel usage associated with various farm equipment (ISU Extension and Outreach, 2001) and based on FST logs of field operations compiled by Pearsons, Chase et al. (Reference Pearsons2023b). Since COMET-Farm does not include emissions from fertilizer and herbicide production, Cool Farm Tool (CFT) was used to calculate CO2 emissions from fertilizer and herbicide production. The second modeling analysis of CFs relied on IPCC Tier 2 methods, which use region-specific emission factors to calculate emissions (IPCC, 2019). Estimates were made using CFT for all categories, with CO2 from fossil fuel combustion estimated based on inputs into the CFT using data on field operations in FST logs instead of manual calculations (see the Supplementary Material for data inputs to CFT).
COMET-Farm modeling (IPCC Tier 3)
Model description.
COMET-Farm is a web-based tool developed by the USDA and Colorado State University that allows users to calculate GHG emissions and carbon sequestration on agricultural lands using cutting-edge methods for quantifying GHG emissions and carbon sequestration (Paustian et al., Reference Paustian2017). The tool is capable of computing emissions of CO2, N2O, and CH4 in CO2-equivalents from all major on-farm sources, as well as determining whether soils act as a carbon source or sink. Users are required to enter data on farm management related to fields, crops, and livestock via a graphical user interface that employs drop-down menus to allow for the specification of pre-defined values for management practices. The tool uses a multi-tiered IPCC Tier 3 modeling approach with spatially explicit data on climate and soils derived from the PRISM climate database and Soil Survey Geographic Database, respectively. The multi-tiered approach requires a combination of user-defined, site-specific, county-level, and some country-specific data. Specifically, COMET-Farm relies on a modified version of the DayCent, a process-based IPCC Tier 3 model, to calculate emissions for N2O and soil C sequestration (Ball et al., Reference Ball2023). The tool allows for complex crop rotations, and users can specify management practices, such as fertilization, for each year over a 23-year period. Once data are entered for the crop rotation over the 2000–2023 period, the main results are simulated using the same crop rotation for the 2024–2034 period based on historical climate data to estimate GHG emissions.
Data entry and processing.
Management data for each cropping system in the FST (CNV, MNR, and LEG) were entered into the COMET-Farm tool, covering the 2000–2023 time period. Inputs included average annual values for crop yields, fertility inputs, herbicide applications, planting dates, and harvesting dates, as well as average date and number of tillage operations based on a 12-year average (2008–2020). Emission estimates for each cropping system were then generated for a simulated period of 2024–2034 using historical climate data to assess long-term emission patterns. Soil C calculations from COMET-Farm were not included as part of the analysis because these results were shown to be inaccurate based on soil carbon data collected independently by the Rodale Institute (Littrell et al., Reference Littrell2021). Additionally, soil C is generally assumed to be at equilibrium and is therefore usually excluded when calculating CFs for cropping systems.
As mentioned in the preceding section, CO2 from fossil fuels was estimated manually using data from Iowa State University extension on fuel usage associated with various farm equipment (ISU Extension and Outreach, 2001) and based on FST logs of field operations compiled by Pearsons, Chase et al. (Reference Pearsons2023b) (see the Supplementary Material). Fuel usage per hectare for each field activity was then multiplied by the total number of hectares in the FST (~5.1 ha) and summed up to obtain the total diesel fuel used for each crop rotation for the entire 12-year period. The total fuel used in each crop rotation for each system was then multiplied by the emission factor for diesel fuel (0.01018 tons CO2-eq per gallon) to obtain the total and annual average fossil fuel emissions from farm operations for each system. If there was not an exact match for the equipment used in the FST and the Iowa State extension database, an analogous piece of equipment from the database was used as a substitute (e.g., substituting ‘cultipack’ with ‘field cultivate—tilled field’).
As mentioned above, COMET-Farm does not include emissions from fertilizer and herbicide production. Therefore, the CFT was used to estimate CO2 emissions from fertilizer and herbicide production using methods discussed in the ‘Data entry and processing’ subsection in the ‘Cool Farm Tool modeling (IPCC Tier 2)’ section.
COMET-Farm estimates were obtained on total annual emissions per year as well as on a per-hectare basis (i.e., area-scaled emissions) following the procedures discussed above. Emission intensities (i.e., yield-scaled emissions) were calculated to provide further comparison between the three production systems. Since maize is the main staple crop in all three systems, yield-scaled emissions were calculated for maize. To compute yield-scaled emissions for maize, the total absolute annual emissions (kg CO2-eq yr−1) across all crops for the entire 5.1 ha of cropland were divided by the total cumulative maize yield in kilograms for the cropped area. The total cumulative maize yield was calculated by multiplying the average annual maize yield (kg ha−1) for the entire 12-year period for each crop rotation and multiplying by the cropped area (5.1 ha). From these calculations, yield-scaled emissions were obtained in units of kg CO2-eq kg−1 maize.
Since the crop rotations in the two organic systems are complex compared with the conventional system and most of the products can be used as livestock feed, yield-scaled emissions on a per megajoule (MJ) of digestible energy basis were calculated for each crop rotation in order to provide a more accurate comparison between the various products of the three systems (i.e., which system had higher/lower emissions per unit of livestock feed energy). To calculate the energy present in each crop, the Colorado State Extension database on Feed Composition for Cattle and Sheep (v. 1.615) was used to obtain estimates of digestible energy (MJ kg−1). Digestible energy present in each crop was then multiplied by yields for each crop in kilograms to obtain values in MJ ha−1 for each production system. Values in MJ ha−1 were subsequently averaged for the entire 12-year crop rotation to obtain mean energy content for each rotation on an annual basis (MJ ha−1 yr−1). Annual energy for the entire rotation was multiplied by the acreage (5.1 ha) to obtain total annual digestible energy for the entire cropped area (MJ yr−1). Yield-scaled emissions were calculated by dividing total absolute annual emissions (kg CO2-eq yr−1) by the total annual energy present in crops (MJ yr−1), resulting in yield-scaled emissions on an MJ basis (kg CO2-eq MJ−1).
Cool Farm Tool modeling (IPCC Tier 2)
Model description.
Similar to COMET-Farm, CFT is an online tool (https://app.coolfarmtool.org/) that allows users to enter farm management data to compute emissions for all categories mentioned above, as well as carbon storage, at the farm or ranch level for a single year. The CFT relies on IPCC Tier 2 emission factors to calculate annual emissions and carbon storage in soils. Much like COMET-Farm, emissions can be obtained for N2O from soils and fertilizer, but the CFT also allows for the calculation of emissions from farm equipment and operations (i.e., fossil CO2), fertilizer production, and pesticide production.
Data entry and processing.
Since the CFT does not allow for complex crop rotations and only allows for one crop and year to be modeled at a time, every individual crop in the rotation was entered separately for all production systems using 2024 as the base year. As described in the ‘COMET-Farm modeling (IPCC Tier 3)’ section, the data entered were based on average annual values for each crop in the rotation for the 12-year period for which field data were available (see the Supplementary Material). Since the CFT can estimate emissions in all categories (except for CO2 from urea fertilization), GHG emissions estimates produced by the CFT were used for each emissions category. Following data entry and extraction of GHG emissions estimates for each crop using the CFT, emissions for each category were summed up together for all crops in the rotation, then divided by the number of years in the rotation to obtain annual averages of GHG emissions (tons CO2-eq yr−1 and tons CO2-eq ha−1 yr−1) for each emissions category for the entire crop rotation. To calculate yield-scaled emissions (kg CO2-eq kg−1 maize and kg CO2-eq MJ−1) from CFT modeling results, the same procedures were followed as described in the ‘Data entry and processing’ subsection in the ‘COMET-Farm modeling (IPCC Tier 3)’ section for the COMET-Farm modeling outputs.
Secondary analysis
Emissions from manure and composting processes are not usually included in LCAs and CF studies of arable cropping systems, such as those found in the FST because these emissions are considered part of the livestock system (i.e., outside the system boundaries) (Venkat, Reference Venkat2012). There is considerable debate in the literature over where emissions from manure should be attributed to LCAs and CFs when manure or compost is applied to croplands (Meier et al., Reference Meier2015). However, preliminary results using the CFT showed that emissions of CO2 associated with the production of composted manure used as a soil amendment were high and had a major impact on the CF of the MNR system. Since the integration of crop and livestock systems is common and represents a major research priority for organic production systems, it is important to assess the effects of including emissions associated with compost production on the overall GHG profile. Therefore, emissions of CO2 were included for compost production as estimated by the CFT and averaged over the entire crop rotation for the MNR system as part of a separately conducted secondary analysis. The CFT computes emissions of CO2 associated with composting but does not provide estimates of N2O and CH4 emissions; therefore, the overall emissions from composting may be underestimated in the present study. Compost CO2 emissions were added to the results of the baseline analysis that was performed using CFT discussed in the ‘Cool Farm Tool modeling (IPCC Tier 2)’ section, and calculations of total absolute, area-scaled, and yield-scaled emissions were conducted using the procedures discussed above for CFT results (subsection ’Data entry and processing’ in the ‘Cool Farm Tool modeling (IPCC Tier 2)’ section).
Results
Baseline analyses
The baseline analysis of area-scaled emissions revealed consistent patterns across modeling approaches for all three systems, though with notable differences in magnitude. Emissions estimated using the COMET-Farm model were, on average, 27% higher than those estimated with the CFT across all systems, with the largest discrepancy observed in the LEG system, where COMET-Farm emissions were 55%–58% higher (Figure 1). Across both models, the CNV system exhibited the highest area-scaled emissions, which ranged from 1.68 to 1.72 tons CO2-eq ha−1 year−1 for the COMET-Farm model and 1.25 to 1.32 tons CO2-eq ha−1 year−1 for the CFT. This was followed by the MNR system with area-scaled emissions ranging from 1.06 to 1.09 tons CO2-eq ha−1 year−1 for COMET-Farm to 0.94 to 0.98 tons CO2-eq ha−1 year−1 for CFT, which resulted in lower emissions by 37% and 25%–27% compared with the CNV system for the two models, respectively. Emissions were lowest in the LEG system for both models and were lower by 52%–54% (0.78–0.83 tons CO2-eq ha−1 year−1) and 73%–74% (0.33–0.37 tons CO2-eq ha−1 year−1) compared with the CNV system for COMET-Farm and CFT, respectively.
Figure 1. Plots display area-scaled annual greenhouse gas emissions in metric tons of CO2-equivalents per hectare broken down by category for the baseline scenario for the COMET-Farm (panel (a)) and Cool Farm Tool (panel (b)) models. Categories are indicated by color and broken down into CO2 from fertilizer application (CO2), N2O emissions from soil/fertilizer (N2O), CO2 from combustion of fossil fuels (fossil CO2), CO2 from fertilizer production (fertilizer production), and CO2 from herbicide production (herbicide production). Treatments are conventional with full tillage (CNV FT) and reduced tillage (CNV RT), organic manure with full tillage (MNR FT) and reduced tillage (MNR RT), and organic legume with full tillage (LEG FT) and reduced tillage (LEG RT).
A similar trend was observed for yield-scaled emissions of maize (Figure 2). For yield-scaled emissions of maize in COMET-Farm, emissions were highest for CNV and similar between the MNR and LEG systems. Compared with the CNV system, yield-scaled emissions were lower by 36%–38% and 35%–40% using COMET-Farm in the MNR and LEG systems, respectively. Yield-scaled emissions for maize in CFT were highest for CNV, followed by the MNR (−26% to −32%) and LEG (−64%) systems. On a per MJ of digestible energy basis, yield-scaled emissions were generally higher in the COMET-Farm model than in CFT (Figure 3). The pattern of yield-scaled emissions among the three systems diverged somewhat between models per unit of digestible energy. The MNR system had the lowest emissions in the COMET-Farm model, whereas yield-scaled emissions were lowest for the LEG system using CFT. For the COMET-Farm model, yield-scaled emissions for digestible energy compared with the CNV system were lower by 25%–26% in the LEG and 47%–50% in the MNR system. For the CFT model, this situation was reversed and yield-scaled emissions were lower compared with CNV by 36% in the MNR system and 57% in the LEG system.
Figure 2. Plots display the plot of yield-scaled greenhouse gas emissions in kg CO2-equivalents per kg of maize for the baseline scenario for the COMET-Farm (panel (a)) and Cool Farm Tool (panel (b)) models. Treatments are conventional with full tillage (CNV FT) and reduced tillage (CNV RT), organic manure with full tillage (MNR FT) and reduced tillage (MNR RT), and organic legume with full tillage (LEG FT) and reduced tillage (LEG RT).
Figure 3. Plots display the plot of yield-scaled greenhouse gas emissions in kg CO2-equivalents per kg of megajoule (MJ) of livestock feed energy for the baseline scenario for the COMET-Farm (panel (a)) and Cool Farm Tool (panel (b)) models. Treatments are conventional with full tillage (CNV FT) and reduced tillage (CNV RT), organic manure with full tillage (MNR FT) and reduced tillage (MNR RT), and organic legume with full tillage (LEG FT) and reduced tillage (LEG RT).
Across all systems and models, N2O emissions from soils and fertilizer were the primary contributor to total GHG emissions. The highest N2O emissions were recorded in the CNV system using COMET-Farm and in the MNR system using CFT. This was followed by emissions from fertilizer production (CNV system only) and CO2 emissions from fossil fuels, the latter of which varied little between production systems (Figure 1). Differences in tillage systems (FT and RT) were consistently smaller compared with differences between production systems and models across all emissions metrics.
Compost emissions secondary analysis
The inclusion of CO2 emissions from the composting process altered the comparative performance of the MNR system relative to the CNV system observed in the baseline analyses (Figure 4). Annual area-scaled emissions under this scenario were highest in the MNR system at 3.25–3.30 tons CO2-eq ha−1 year−1 and CO2 emissions from compost production (that is fertilizer) were the main driver of emissions for this system. Incorporating composting emissions shifted the MNR system from a lower-emission option to one with higher overall GHG emissions compared to both the CNV and LEG systems. This trend was consistent across different metrics, as the MNR system showed elevated emissions on a yield-scaled basis for both mass (e.g., kg CO2-eq per kg of maize) and energy (e.g., kg CO2-eq per MJ of digestible energy). These findings underscore the significant impact that including emissions from compost production can have on the overall CF of organic systems, highlighting the importance of system boundary considerations in CF assessments.
Figure 4. Plot displays area-scaled annual greenhouse gas (GHG) emissions in metric tons of CO2-equivalents per hectare broken down by category for the secondary analysis that includes emissions associated with the production of composted manure. Categories are indicated by color and broken down into CO2 from fertilizer application (CO2), N2O emissions from soil/fertilizer (N2O), CO2 from combustion of fossil fuels (fossil CO2), CO2 from fertilizer production (fertilizer production), and CO2 from herbicide production (herbicide production). Treatments are conventional with full tillage (CNV FT) and reduced tillage (CNV RT), organic manure with full tillage (MNR FT) and reduced tillage (MNR RT), and organic legume with full tillage (LEG FT) and reduced tillage (LEG RT).
Discussion
As initially hypothesized, absolute area-scaled emissions were lower for organic systems compared with conventional systems, and the magnitude of these differences was broadly in line with those reported in the literature (Reganold and Wachter, Reference Reganold and Wachter2016; Seufert and Ramankutty, Reference Seufert and Ramankutty2017). One review of the literature found that area-scaled emissions for cereals were 39% lower in organic systems than conventional systems, which compared well with the 25%–74% decrease in area-scaled emissions for organic across all systems and models in the present study (Chiriacò, Castaldi and Valentini, Reference Chiriacò, Castaldi and Valentini2022). Work on long-term trials similar to the FST in Switzerland found that area-scaled N2O emissions were ~40% lower for the organic system compared with the conventional system (Skinner et al., Reference Skinner2019). By contrast, a meta-analysis conducted on measured N2O emissions from arable crops found that area-scaled emissions were somewhat higher in organic systems at 1.51 tons CO2-eq ha−1 year−1 compared with 1.39 tons CO2-eq ha−1 year−1 for conventional systems (Skinner et al., Reference Skinner2014). Another study measured GHG emissions from one organic versus two conventional rotations of maize-soybean-wheat and reported 0.41 tons CO2-eq ha−1 year−1 for the organic system compared with 0.14–1.14 tons CO2-eq ha−1 year−1 in the conventional systems (Robertson, Paul and Harwood, Reference Robertson, Paul and Harwood2000). The lower average area-scaled emissions for conventional systems in the aforementioned study were mainly due to the inclusion of soil carbon sequestration under no-till management (the present study did not include an analysis of soil carbon sequestration). Overall, these previously reported results compared well with our range of 0.33–1.09 tons CO2-eq ha−1 year−1 for the organic systems but were lower than results obtained here at 1.25–1.72 tons CO2-eq ha−1 year−1 for the CNV system.
N2O emissions from fertilizer applications and soils were the largest source of emissions in all systems in the baseline analyses. N2O emissions surpassed those from all other sources combined, consistent with other studies that have identified N2O as the primary driver of emissions in maize-based systems in the United States (Pelletier, Arsenault and Tyedmers, Reference Pelletier, Arsenault and Tyedmers2008; Kim, Dale and Keck, Reference Kim, Dale and Keck2014). Similarly, research comparing conventional and organic maize–wheat–soybean rotations found that N2O was the dominant driver of emissions in all systems, regardless of N source (Hoffman et al., Reference Hoffman2018). Our findings differ from studies on cereals and legumes in Mediterranean climates, which found that CO2 emissions from fossil fuel use were the primary driver of total emissions. This difference is likely due to the low N2O emission factor in Mediterranean regions, driven by low rainfall (Aguilera, Guzmán and Alonso, Reference Aguilera, Guzmán and Alonso2015).
Rates of N application in both the CNV and MNR systems of the FST align with typical cropping practices in eastern Pennsylvania. Fertilizer rates in CNV follow recommendations from Pennsylvania State University Extension, while the MNR system replicates a 54-ha dairy farm (the average farm size in Pennsylvania) that returns manure back to the fields. Despite the composting of manure prior to application, N2O emissions in the organic MNR system were comparable to emissions from synthetic fertilizer in the CNV system in both models. This supports recent research suggesting that the quantity of N applied is a more critical factor than its source in driving GHG emissions (Han, Walter and Drinkwater, Reference Han, Walter and Drinkwater2017). Recent research has also shown that applying manure or compost at the same time as cover crops may result in synergistically higher emissions than either of the two practices alone (Saha et al., Reference Saha2021).
In terms of differences between models, we found that COMET-Farm generally had higher absolute area-scaled emissions than CFT, particularly for the LEG system. The divergence between models was largely driven by differences in N2O emissions since COMET-Farm reported higher N2O emissions compared with CFT. Differences in N2O emissions between models are likely driven by differences in how the models estimate N2O emissions from soils and fertilizers. COMET-Farm uses a processed-based model to simulate N2O emission dynamically based on geographically specific parameters related to soil properties (e.g., soil type and soil fertility level), local climatic conditions, and user-specified fertility inputs and management practices. According to the IPCC guidelines, COMET-Farm can be classified as a Tier 3 model for estimating emissions, which are generally considered the most complex and accurate models available, provided that adequate data are available for model parameterization (IPCC, 2019). On the other hand, CFT relies mainly on Tier 2 emission factors, which are intermediate in terms of complexity and generally considered somewhat less accurate than Tier 3 models due to the lack of location-specific data used for estimating emissions. These differences in methodologies between the COMET-Farm and CFT models likely explain the differences in emissions between the two models, and the COMET-Farm model should be regarded as more likely to produce accurate estimates of N2O emissions in this instance.
Another source of variation between the models may stem from the exclusion of cover crops in the crop rotation in the CFT modeling since this was not possible within the CFT framework. Although N2O emissions from cover crops tend to be lower than from other sources, such as manure and synthetic fertilizers, cover crop N2O emissions are complex and depend on whether the cover crop is leguminous or non-leguminous, as well as whether the cover crop is incorporated into the soil or not (Muhammad et al., Reference Muhammad2019). Leguminous cover crops tend to increase N2O emissions due to their relatively high N content, especially when cover crop residues are incorporated, whereas non-leguminous cover crops may reduce emissions because they can scavenge mineral N from the soil and tend to have higher C:N ratios (Basche et al., Reference Basche2014). Since N2O emissions from cover crops were not captured as part of the CFT modeling, the CFT may underestimate overall N2O emissions compared to the COMET-Farm tool. On the other hand, it is unclear how well COMET-Farm and the underlying DayCent model can simulate emissions from cover crops since, critically, COMET-Farm does not require inputs for biomass quantity or N content of the cover crop when computing emissions. Some research has shown that the CENTURY model, on which DayCent and COMET-Farm are based, tends to underestimate emissions from cover crops compared to field measurements (Xia et al., Reference Xia2022).
Research highlights that current modeling approaches may miscalculate N2O emissions from different nitrogen sources, pointing to a need for further studies on this topic (Xia et al., Reference Xia2022; Xia and Wander, Reference Xia and Wander2022). Verifying the results from this study with in-situ measurements of N2O emissions in the FST, particularly those comparing composted manure to synthetic fertilizers, is critical for validating model estimates.
Contrary to our original hypothesis and what has previously been reported in the literature, yield-scaled emissions were lower for organic systems than conventional systems on a per unit of output basis (i.e., per kg of maize and per MJ). In the MNR system, average maize yields (6,717–6,993 kg ha−1) closely matched those in the CNV system (6,296–6,897 kg ha−1), contributing to the lower yield-scaled emissions in MNR system. Although average maize yields were lower by ~20% in the organic LEG system (4,582–5,597 kg ha−1) compared with CNV, this system’s substantially lower area-scaled emissions led to reduced yield-scaled emissions overall.
These results contrasted with those reported in a literature review, which found that yield-scaled emissions for cereal crops were ~8% higher for organic systems, though this was based on relatively few studies (n = 3). Our results also contrasted with a meta-analysis showing that organic cereals had higher yield-scaled emissions compared with conventional cereals (Clark and Tilman, Reference Clark and Tilman2017). There was only one study that directly compared organic and conventional systems with maize as the main staple crop and reported yield-scaled emissions, and our results largely aligned with those reported in this previous study of lower emissions in the organic system (0.26 kg CO2-eq kg−1) compared to the conventional system (0.35 kg CO2-eq kg−1) (Pelletier, Arsenault and Tyedmers, Reference Pelletier, Arsenault and Tyedmers2008). This compared well to the ranges found in the present study of 0.19–0.25 kg CO2-eq kg−1 maize for CNV systems and 0.07–0.17 kg CO2-eq kg−1 maize for the organic systems (Figure 2). The range of yield-scaled emissions for conventionally grown maize in the literature was between −0.03 and 0.44 kg CO2-eq kg−1 maize (the study with a negative value included soil carbon sequestration as part of the CF, leading to negative emissions) (Kim, Dale and Keck, Reference Kim, Dale and Keck2014). These figures give us broad confidence that our estimates are within the range reported for maize on a yield-scaled basis in the literature, but this is still based on a limited number of studies. There remains a need to conduct more direct comparisons between conventional and organic systems to inform meta-analysis and large-scale modeling efforts, especially since modeling often predicts net increases in GHG emissions due to broad-scale conversion to organic agriculture (Smith et al., Reference Smith2019).
Yield-scaled emissions on a per MJ basis largely followed the pattern found for maize, albeit with a closer match between the MNR and LEG systems (Figure 3). The crop rotations in both organic systems, particularly the MNR system, are complex and focusing exclusively on maize yield-scaled emissions could obscure differences in emissions between crops in the same rotation, leading to under- or overestimates of emission intensity (Costa et al., Reference Costa2020). It may also omit the fact that either the organic or conventional system could be producing more total crop output and therefore lead to mistaken conclusions about system performance. It has been suggested in the literature to use the energy present in crops, or, in the case of LCAs, use a different functional unit such as the recently developed ‘cereal unit’, which allows for comparison of impacts across varying products in a crop rotation (Brankatschk and Finkbeiner, Reference Brankatschk and Finkbeiner2014). The relatively close correspondence between maize-based and MJ-based estimates of yield-scaled emissions implies that our results are reasonably robust and validate the presence of reduced yield-scaled emissions in organic systems compared to conventional.
Including compost CO2 emissions in the secondary analysis appears to move the MNR system from a ‘climate winner’ in the baseline analyses to a ‘climate loser’. This outcome illustrates the sensitivity of system performance to boundary conditions and highlights the importance of including all relevant emissions in LCA and CF studies, especially in integrated crop–livestock systems. It suggests that organic systems relying on manure may not always achieve lower GHG emissions, depending on assumptions about system boundaries (Rehberger et al., Reference Rehberger2023). However, the use of organic-based fertility sources increases soil carbon and other measurements of soil health (Hepperly et al., Reference Hepperly, Lotter, Ulsh, Seidel and Reider2009; Bagnall et al., Reference Bagnall2022; Liptzin et al., Reference Liptzin2022), and some evidence has shown that grazing livestock has the potential to sequester more soil carbon than is emitted on an equivalency basis from enteric CH4 and soil GHG emissions (Rowntree et al., Reference Rowntree, Ryals, DeLonge, Teague, Chiavegato, Byck, Wang and Xu2016; Stanley et al., Reference Stanley, Rowntree, Beede, DeLonge and Hamm2018; Rowntree et al., Reference Rowntree, Stanley, Beede, DeLonge and Hamm2019).
As mentioned previously, there is considerable debate about how emissions from manure and manure-derived products, such as compost, should be attributed to CF studies. Indeed, some of the literature on this topic points out that instead of being penalized for emissions from compost, integrated crop–livestock systems ought to be rewarded with a GHG ‘credit’ because they do not produce GHGs associated with fertilizer production (i.e., fossil CO2 emissions as reported here) (Meier et al., Reference Meier2015). This is particularly salient, given that composting tends to reduce overall emissions from manure compared to other manure management systems and is thus considered a climate-smart agricultural practice (Petersen et al., Reference Petersen2013; Vergara and Silver, Reference Vergara and Silver2019). The CFT calculates only CO2 emissions due to compost and does not account for N2O and CH4 emissions, potentially leading to underestimates of overall GHG emissions from composting. However, CO2 emissions from composting are considered biogenic and therefore do not contribute to global warming (Nordahl et al., Reference Nordahl2023), so it is unclear whether these should be included as part of the overall CF for systems in which compost is applied. To account for these considerations, a more holistic approach is needed: one that compares integrated organic crop–livestock systems against a conventional system without crop–livestock integration to uncover the true potential for integrated systems to reduce CFs from agriculture. To this point, there has been only limited research regarding the potential climate change mitigation benefits of integrated crop–livestock systems (dos Reis et al., Reference dos Reis2021; van Selm et al., Reference van Selm2023; Delandmeter et al., Reference Delandmeter2024).
The organic LEG system was unaffected by the compost emissions analysis and remained the best-performing system across all analyses, apart from yield-scaled emissions on a per MJ basis using the COMET-Farm model. This could indicate that rotations relying more on legumes and cover crops may be more sustainable than those relying on animal agriculture for soil fertility. This occurred even though the LEG system had ~20% lower yields compared with the CNV system. Moreover, recent research has shown that drastic reductions in food system emissions are required in achieve international goals of keeping global warming to ‘well-below’ 2°C (i.e., 1.5°C) as prescribed by the 2015 Paris Agreement (Clark et al., Reference Clark2020). Most of the aforementioned research has shown that dramatic reductions in livestock emissions, particularly from enteric CH4, and a shift toward more plant-based diets will be required to meet these targets (Muller et al., Reference Muller2017; Arndt et al., Reference Arndt2022). Taken together, this could imply that the organic LEG system, in addition to consistently performing best in this study, may also be the most sustainable system in a world where the dependence on animal agriculture as a source of fertility may need to be reduced.
Finally, our study did not conduct a true LCA of the systems present in the FST to compare CFs and instead used a combination of different methods. Since LCAs are the gold standard for estimating emissions from agricultural systems and other products, future research should examine the possibility of conducting a true LCA on the FST to determine associated CFs. Although soil carbon sequestration was not included as part of the analyses in this study, the two organic systems have higher carbon stocks compared with the conventional system (Littrell et al., Reference Littrell2021; Pearson, Omondi et al., Reference Pearson2023a). Soil C in the FST has changed over time as a result of management (Drinkwater, Wagoner and Sarrantonio, Reference Drinkwater, Wagoner and Sarrantonio1998), with the MNR system having the highest carbon stocks. Furthermore, the use of compost as a fertility source has the potential to increase soil carbon stocks more than synthetic fertilizer and uncomposted manure (Hepperly et al., Reference Hepperly, Lotter, Ulsh, Seidel and Reider2009), and thus future research of these systems should include carbon sequestration as part of CF modeling. Our results are largely based on data from field operations and not measurements of GHG emissions from the field. Future research should conduct in-situ field measurements to verify these results, particularly regarding N2O emissions from these systems.
Conclusions
Overall, these results show that organic agricultural systems can outperform conventional systems when it comes to their CFs and GHG emissions on an absolute, area-scaled, and yield-scaled basis, implying that a shift toward organic farming could facilitate climate change mitigation. Since emissions were lower across all emission metrics and models, these results appear to be relatively robust. Outcomes were sensitive to model choice, modeling assumptions, and system boundaries, with the IPCC Tier 3 COMET-Farm approach resulting in higher emissions compared with CFT (IPCC Tier 2). Emissions were sensitive to systems boundaries when dealing with integrated crop–livestock systems in the secondary analyses on composting, and it will be important to consider whether and how emissions associated with animal agriculture ought to be included within the system boundaries when conducting CF analyses and LCAs. Future research should attempt to better capture the effects of complex crop rotations in organic systems on CFs using LCA methods to inform broad-scale modeling of organic agriculture and verify these results using in-situ field measurements of major GHGs.
Supplementary material
The supplementary material for this article can be found at http://doi.org/10.1017/S1742170525100069.
Acknowledgements
We would like to acknowledge the farm operations team, researchers, technicians, and interns that have kept FST running for the past 40 years, including farm managers, research staff, and research interns.
Funding statement
This material is based on work supported by the William Penn Foundation under Grant Award Number 188-17. Research and writing for this publication were supported by funding from the Richard King Mellon Foundation, Grant ID # 13496. The opinions expressed in this publication are those of the authors and do not necessarily reflect the views of the William Penn Foundation or Richard King Mellon Foundation.
Competing interests
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.
Open access
