|
Assessing the Feasibility of Meeting Tolerance Limits for GM Adventitious Presence in Corn Supply Chain Using Probabilistic Modeling
Priyanka Gupta1,2, Charles R. Hurburgh1,*, Erin Bowers1, Gretchen A. Mosher1
Published in Applied Engineering in Agriculture 39(5): 543-551 (doi: 10.13031/aea.15570). Copyright 2023 American Society of Agricultural and Biological Engineers.
1Department of Agricultural and Biosystems Engineering, Iowa State University, Ames, Iowa, USA.
2Department of Technology, Missouri Southern State University, Joplin, Missouri, USA.
*Correspondence: tatry@iastate.edu
The authors have paid for open access for this article. This work is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License https://creative commons.org/licenses/by-nc-nd/4.0/
Submitted for review on 21 February 2023 as manuscript number PRS 15570; approved for publication as a Research Article by Associate Editor Dr. Mark Casada and Community Editor Dr. Sudhagar Mani of the Processing Systems Community of ASABE on 28 July 2023.
Highlights
Abstract. Tolerance limits for the adventitious presence (AP) of GM material in non-GM grain, food, and feed vary worldwide from 0.9% to 5.0%. This research analyzed the likelihood of meeting four common trade tolerance limits for AP (0.9%, 1.5%, 3.0%, and 5.0%) in the U.S. commodity corn supply chain. A model was developed to evaluate existing practices and patterns for bulk corn production, handling, and processing in an open-market supply chain that concurrently handles GM and non-GM products. Monte Carlo simulation was used to test 50,000 iterations of supply chain scenarios to determine the likelihood of successfully meeting specified tolerance limits. The model revealed that the supply chain, as it exists today, does not effectively facilitate the concurrent handling of GM and non-GM streams at 0.9% and 1.5% tolerance limits in most cases. At individual supply chain stages, some tolerance limits were reasonably achievable, such as 3.0% and 5.0% at the farm stage. The probabilities of complying with 0.9% and 1.5% tolerance limits at the farm stage were just over 10% and 67%, respectively, while the probabilities of complying with 3.0% and 5.0% tolerance limits were more than 90%. The grain elevator and grain processor could achieve 3.0% and 5.0% tolerance limits with reasonable likelihood. At the feed mill, a 5.0% tolerance limit was achievable but only when bypassing some supply chain stages. The 99% feasible tolerance limits at individual supply chain stages ranged from 2.25% to 6.25%. Significant factors influencing the ability to meet AP tolerances were identified using sensitivity analysis. These factors included seed impurity, cross-pollination, and transportation vehicles.
Adventitious presence (AP) refers to the unintentional presence or accidental commingling of a trace amount of genetically modified (GM) material in non-GM seed, grain, food, or feed products (USDA-AC21, 2005). Many countries have implemented policies that specify tolerance limits for the adventitious presence of GM material in non-GM products. In the United States (U.S.), the National Bioengineered Food Disclosure Standard specifies a tolerance limit of 5.0% AP (per ingredient) in some human foods. Exceeding this limit requires a mandatory GM disclosure to U.S. consumers (USDA-AMS, 2018). The European Union (EU) has the most stringent tolerance limit of 0.9% for EU-authorized GM AP in food and feed (European Commission, 2003). Japan specifies a tolerance limit of 5.0%, and South Korea has a tolerance limit of 3.0%. These countries are among the top export destinations for U.S. corn, soybeans, and other derived feed ingredients (AFIA, 2021; USDA-FAS, 2021).
Currently, the majority of U.S. corn (90%) and soybeans (94%) are genetically modified to confer traits such as insect resistance and herbicide tolerance (USDA-ERS, 2020). A relatively small proportion of non-GM grain is produced, mainly under the contract production mechanism. Non-GM farmers generally contract with grain buyers, who establish specific requirements such as precise tolerances for AP, crop growing and segregation practices, schedules for delivery, and premiums if contract specifications are met (USDA-AC21, 2016). Another commonly used mechanism is containerized shipments, in which specialty grain is marketed under the container and bag systems (Elbehri, 2007). These mechanisms function in closed systems and are well-suited for high-value crops such as food-grade corn and soybeans. If the market demand for non-GM grain expands for high-volume products, such as non-GM feed, premiums would need to be sufficient to justify contract or containerized mechanisms. Otherwise, bulk grain production and handling infrastructure will need to be utilized to support this growth. In these instances, the existing commodity-based infrastructures, historically driven by and designed for high volumes and low-cost handling, will be required to accommodate non-GM crops to meet non-GM demand.
The bulk grain and feed supply chain, as it exists in its current state, is less adept at handling non-GM grain. Grain elevators typically store and mix grains of different qualities to meet grade standards and maximize profitability (Laux et al., 2015). The current grading system relies primarily on tests for physical attributes such as foreign material and damage, in addition to measurements of test weight and moisture content. It is uncommon to test incoming grain for specialty traits such as GM content unless specified in the contract. Segregating grain based on GM content is not a usual practice in these systems. Recent developments in the non-GM market and the tightening of national and international labeling laws may impact the bulk grain industry to adopt grading and segregating systems based on GM traits.
Concerns with segregating GM and non-GM grain at the farm and downstream supply chain stages include the risk of introducing AP. Several factors can potentially contribute to AP at each stage, such as seed impurity, on-farm cross-pollination between neighboring GM and non-GM cornfields, inadvertently spilled or carried GM grain by machinery, and commingling during handling, storing, and processing operations. Detection of GM content in non-GM lots above tolerance limits may lead to a lack of premium payment and losses if found in non-GM markets. Cases have been reported in the past where GM AP in non-GM shipments have impacted domestic and international trade, such as the StarLinkTM situation of 2000 (Harl et al., 2001). Another example is the AP of Bt10 corn, which was unapproved for any end use in 2005, which led to economic losses to its developer company Syngenta and an EU ban on the import of some corn products from the U.S. (Syngenta International AG, 2005).
Prior studies have assessed the impact of individual factors on the feasibility of achieving specified AP limits (Ingles et al., 2003, 2006; Devos et al., 2005; Pla et al., 2006; Weber et al., 2007; Dolphin et al., 2020; Hanna and Jarboe, 2011). Little assessment has been completed on how different factors collectively impact overall tolerance limits in an entire supply chain. Dolphin et al. (2020) examined the collective impact of four factors from a farm-to-elevator (one-step) supply chain: seed impurity, field isolation distance, combine cleanout, and grain elevator receipt and handling practices. Using the Monte Carlo simulation, Dolphin et al. (2020) evaluated nine scenarios to determine the feasibility of achieving 0.9%, 1.5%, and 3.0% tolerance limits in non-GM corn loads and identified that 0.9% tolerance limit was not feasible in most scenarios, and 1.5% and 3.0% were feasible in certain scenarios.
Monte Carlo simulation methods are effective for probabilistic risk assessment and decision-making in the food and agricultural sector (Hu et al., 2020). When evaluating AP in the supply chain, a deterministic approach can also be applied, where a point estimate of AP contributed by each factor is used to calculate outcomes. However, there are certain limitations to consider with this approach. One such limitation is the availability of limited data to account for point estimates of AP possible due to each factor, as obtaining such data can be time-consuming and expensive. Furthermore, a deterministic model does not account for variability (e.g., heterogeneity among handlers in the population) and uncertainty (e.g., lack of knowledge about true value). In contrast, probabilistic modeling offers a more comprehensive approach. It represents input data using probabilistic distributions, making it possible to account for both variability and uncertainty (Kruizinga et al., 2008).
The objectives of this study were: (1) to develop a model measuring AP-contributing factors in the bulk grain and feed supply chain using corn production, handling, and processing practices as the test case, (2) to estimate the probability of achieving tolerance limits for AP of 0.9%, 1.5%, 3.0%, and 5.0% in the entire supply chain and different stages individually, (3) to identify factors with the highest influence on AP levels at each supply chain stage.
Materials and Methods
Model Assumptions, Input Data, and Distributions
The model was created using the @Risk software add-in for Microsoft Excel (Palisade Corporation, Ithaca, N.Y.). Actions within the commodity corn supply chain were modeled by dividing it into four stages: farm, grain elevator, grain processor, and feed mill. In figure 1, the first column illustrates the supply chain processes evaluated. The second column summarizes several AP contributing factors that have been identified as important contributors to AP during corn production, handling, and processing in previous studies (Hurburgh, 2000; Ingles et al., 2003, 2006; Maier, 2006; Devos et al., 2009; Mosher and Hurburgh, 2010; Gupta et al., 2022), these factors are discussed in detail in the subsequent sections. To account for the variable levels of AP among handlers, each of the factors was assigned not just a single point value but a probabilistic distribution, as summarized in the third column in figure 1. This enabled stochastic modeling of AP levels in the supply chain, which provides a more realistic representation and models variability more accurately than static modeling (Vose, 2008). Input distributions and values were assigned to each factor by referring to quantitative data from previous studies. Input values are summarized in the fourth column in figure 1 and discussed in detail in the subsequent sections. The model assumed that the amount of AP contributed by the factors aggregate across the supply chain to cause a higher cumulative AP in the final product. Some AP-contributing factors were similar at different supply chain stages due to the similar nature of some activities, such as transportation, grain unloading, and conveying.
 |
| Figure 1. Commodity corn supply chain stages, factors contributing to the adventitious presence (AP) of genetically modified (GM) content in non-GM lots, probabilistic distribution type, and AP (%) mean and standard deviation used as input data for the Monte Carlo simulations in this study. |
The study applied the Monte Carlo method to simulate scenarios by performing repeated random sampling and statistical analysis. The Monte Carlo simulation sampled a random value from input distributions, aggregated them together, and yielded a final output value. This output value reflects one iteration and represents one possible outcome. The study repeated the process for 50,000 iterations generating 50,000 output data points to obtain a comprehensive probabilistic distribution. These output distributions were used to estimate the probability of achieving AP tolerance limits of 0.9%, 1.5%, 3.0%, and 5.0%. The choice of 50,000 iterations in this study aligns with established practices in the field. Previous studies, such as Dolphin et al. (2020), have employed similar iteration counts in Monte Carlo simulations. Using 50,000 iterations in a Monte Carlo simulation offers a reasonable balance between statistical accuracy and computational efficiency.
Different supply chain configurations were created to evaluate the impact of the supply chain structure on its feasibility of achieving specified tolerance limits, discussed in subsequent sections.
Farm Stage
The model analyzed AP at the farm stage considering the following operations for corn production: seed receiving, crop planting, grain harvesting, storage, and transportation. The model analyzed the amount of AP that could occur due to the following factors: seed impurity, cross-pollination, and transportation vehicles. AP due to on-farm grain handling, drying, and storage were not included in the model due to the lack of available data.
The first AP-contributing factor evaluated at this stage was seed impurity (GM impurity in non-GM seed lots). A certified bag of non-GM seed cannot be assumed to have zero percent AP (Bradford, 2006). Data on seed impurity was derived by referring to seed purity standards set by international and national standard-setting organizations. For example, according to the Organization for Economics Co-operations and Development (OECD), the maximum limit for the seed of other varieties or off-types is 1% in certified corn seeds and 0.5% in basic (foundation) seeds (OECD, 2021). The model in this study simulated the possible levels of seed impurity ranging from 0% to 1.0% with an average level of 0.5% by weight. As per the federal Seed Act, seed companies are required to label purity data on seed tags that include percent inert material, percent other crop seeds, and percent weed seeds; the Act does not require tags to specify GM presence or percentage (USDA-AC21, 2016). This causes uncertainty about the actual level of GM impurity in a seed bag, which cannot be known until tested. For example, if a bag is labeled with a maximum impurity limit of 0.5% by weight, the actual GM impurity level in the bag can range anywhere from 0% to 0.5%. To represent this variability in this study and obtain a realistic estimation of AP at the farm, the factor seed impurity was modeled as a beta distribution. The beta distribution is often used to model random variables, and in this case, seed impurity is considered an event with a random chance of occurrence (Vose, 2008; Dolphin et al., 2020).
The next AP-contributing factor evaluated was cross-pollination between GM and non-GM cornfields. Previous studies have shown that cross-pollination can occur up to 500 m between GM and non-GM fields (Devos et al., 2005), therefore, AP due to cross-pollination cannot be assumed to be zero percent. A study by Dolphin et al. (2020) evaluated the impact of three isolation distance ranges of ‘0-20 m,’ ‘21-40 m,’ and ‘41 m+’ on the feasibility of achieving AP tolerance limits and found that the feasibility to achieve AP limits such as 0.9% increased substantially with isolation distance. Referring to the findings of Dolphin et al. (2020), the current study simulated AP due to cross-pollination using a Weibull distribution with an average level of 0.39%. The study assumed that the farmer implemented an isolation distance ranging between 20-40 m by referring to previous studies suggesting that around 95% to 99% of the pollen released is typically deposited within a radius of 30 m from the source (Devos et al., 2005). Weibull distribution was applied to represent the non-random pattern observed in the isolation distance data (Dolphin et al., 2020), as it is often used to characterize non-random point clouds or distributions.
Transportation vehicles are also a factor. If the cleanout is absent or insufficient, the transportation vehicle may carry over residual GM grain from previous loads and commingle it with non-GM loads, contributing to AP. The level of AP contributed by transportation vehicles was modeled using a triangular distribution to simulate three possible values: 1% for the maximum level, 0.3% for the most likely level, and 0.01% for the minimum level, by referring to data by Wilson et al. (2003).
Grain Elevator Stage
The model analyzed AP at the country elevator with general operations: grain receiving, unloading and conveying, grain storage, and transportation. The AP-contributing factors evaluated at this stage included grain impurity, unloading and conveying equipment residue, and transportation vehicles. Grain impurity depicted the level of AP in the incoming corn lots due to on-farm operations. The model assumed AP aggregated in corn lots at the farm stage was passed on to the elevator stage as ‘grain impurity.’ The implications of GM testing at the time of receiving were not considered in the model. The model, at this stage, examined the amount of AP that could be added to grain impurity due to operations at the grain elevator.
The model assumed that the grain elevator handled both GM and non-GM corn using the same flow path and equipment for the loads. Unloading and conveying operations at the grain elevator included the dumping of grain into the receiving pit and then conveying it to the storage bin via equipment such as conveyors, bucket elevators (leg), distributors, weighing scales, and grain cleaners. Data from Ingles et al. (2006) was referenced to derive the level of AP that might commingle in the unloading and conveying system if elevators are run nonstop without cleaning between GM and non-GM loads. According to the study, grain flow through the pit, leg, and drag conveyors could result in a cumulative commingling of 0.3%. The level of commingling might decrease as the volume of grain increases (Boac et al., 2012), and therefore, an exponential distribution was used to represent AP due to unloading and conveying equipment in the model.
The grain is then transported to a domestic grain processor or feed mill, or river terminal for overseas export. Due to the lack of available data, AP contributed by transportation vehicles from the grain elevator to the grain processor or feed mill was assumed to be similar in distribution to that from the farm to the grain elevator in this study.
Grain Processor Stage
The grain processor stage represents the conversion of corn into co-products used as feed ingredients. The model analyzed AP at a grain processor with general operations: receiving, unloading, conveying, processing, final product storage, and transportation to a feed mill. The model estimated the amount of AP that the following factors could contribute - grain impurity, unloading and conveying equipment residue, processing equipment residue, and transportation vehicles. The grain impurity factor at this stage represents the aggregated level of AP that is passed on from the farm and elevator to the processor.
The grain unloading and conveying operations were assumed to contribute similar amounts of AP as that at the grain elevator stage. Processing equipment such as grinders, mixers, fermenters, and dryers may accumulate and carry over residual GM material to non-GM batches. Sufficient data was not available to simulate the levels of AP contributed by each piece of equipment individually. The overall AP contributed by the processing line was simulated using an exponential distribution. The exponential distribution accounts for a high level of AP at the start of the operation (as if GM grain had just been handled) and declines as increasing volumes of material (in this case, non-GM grain) pass through the line. The average level of AP contributed by a processing line was 0.6%, derived by referring to the study by Ingles et al. (2003). AP contributed by transportation vehicles from the grain processor to the feed mill was assumed to be similar to that from the farm to the grain elevator in this study.
Feed Mill Stage
The model analyzed AP at a commercial feed mill that produces swine and poultry feed using corn arriving from the farm and grain elevator stages and feed-ingredient arriving from the grain processor stage. General operations at the feed mill start with the ingredient receiving, unloading and conveying, grain milling, ingredient mixing, feed pelleting, feed bagging, storage, and transportation. The model analyzed the amount of AP that could occur due to the following factors: ingredient impurity, unloading and conveying equipment residue, processing equipment residue, and transportation vehicle.Ingredient impurity represents the aggregated level of AP that passes from the upstream supply chain participants (farm, elevator, processor) to the feed mill.
The grain unloading and conveying operations were assumed to be similar to those at the elevator and processor and were assumed to contribute to similar amounts of AP as those of the upstream supply chain stages, due to the lack of available data specific to feed processing. Processing equipment for swine and poultry feed include milling systems that grind whole corn, batching and mixing systems to mix ingredients to form a mash or meal feed, conditioning and pelleting systems to form pelleted feed, drying and cooling systems to reduce moisture and temperature, crumbling systems to form crumbled feed, and bagging systems to package feed. Processing equipment such as grinders, mixers, pellet mills, coolers, and dryers potentially accumulate GM material and commingle non-GM batches. Sufficient data were not available to simulate AP levels contributed by each piece of equipment individually. The overall AP contributed by a feed processing line was simulated using an exponential distribution. The average level of AP contributed by processing equipment was assumed to be 0.6%, derived by referring to the data by Ingles et al. (2003). AP contributed by transportation and distribution vehicles was assumed to be similar to that at the upstream stages.
Supply Chain Configurations
The model was created to evaluate AP and changes in AP at each supply chain stage as well as stages combined. Different supply chain configurations were created to evaluate the impact of the supply chain structure on its feasibility of achieving specified tolerance limits. Three configurations were created based on the combination of different stages. In configuration 1, all stages were present in the supply chain, signifying the flow of ingredients such as corn and feed ingredients to the feed mill from the grain processor, grain elevator, and farm. In configuration 2, the grain processor stage was bypassed; the feed mill received feed ingredients, such as corn, from the farm via an elevator. In configuration 3, the grain processor and grain elevator stages were bypassed, and the feed mill received grain directly from the farm.
Sensitivity Analysis
After running the model simulations, a sensitivity analysis was conducted to identify factors that significantly impacted model outputs. Variance-based sensitivity analysis was performed using the ‘tornado-contribution to variance’ feature of @Risk software. The analysis allowed the breakdown of the model output variance into fractions, which could be attributed to a selected number of input factors. This allowed the identification of inputs that had the highest impact on output variance. Identification of significant contributors to output variance has useful implications as it allows the identification of opportunities for improvement. It makes it possible to target resources to a small number of significant factors to attain low tolerance limits rather than distributing resources across all factors.
Results and Discussion
AP levels were modeled stochastically in the grain and feed supply chain with corn production, handling, and processing practices currently used in the supply chain. Monte Carlo simulations were performed to determine the probability of achieving tolerance limits of 0.9%, 1.5%, 3.0%, and 5.0% both at individual supply chain stages and stages combined. Figure 2 shows the output distribution generated after simulating 50,000 iterations for individual supply chain stages. The curved lines represent the probability of achieving AP tolerance limits at each stage. For example, the probability of achieving a 0.9% tolerance limit was 12.2% at the farm stage, 0.1% at the grain elevator stage, and 0% at the grain processor and feed mill stages. Bars in the graph were analyzed to identify tolerance limits achievable with 99% feasibility (the AP level within which 49,500 outputs out of 50,000 iterations fall). At the farm stage, the 99% feasible tolerance limit was 2.25%; at the grain elevator stage, it was 3.10%; at the grain processor stage, it was 4.75%; and at the feed mill stage, it was 6.25%.
Table 1 summarizes the simulation outcomes in this study, including the probabilities with which varying tolerance limits can be achieved at different supply chain stages and configurations, 99% feasible tolerance limits, and outputs of sensitivity analyses. Findings suggest that, given the current practices and factors contributing to AP in a typical commodity corn supply chain, tolerance limits such as 0.9% and 1.5% were not achievable at any supply chain stages. The highest probabilities to achieve 0.9% and 1.5% were just over 10% and 67%, respectively, at the farm stage; at other stages, the probability was near 0%. The farm stage could achieve 3.0% and 5.0% tolerance limits with probabilities of more than 90%. The grain elevator and grain processor could achieve 3.0% and 5.0% tolerance limits with reasonable likelihood, although the model indicated slight probabilities of non-compliance for these stages at these tolerance limits. The feed mill could not achieve any of the specified tolerance limits without a significant likelihood of non-compliance. These findings can be true for broad-scale real-world cases, considering the large number of iterations performed in the study; the iterations randomly sampled so many inputs that provided an understanding, on the whole, of what a typical, present-day supply chain stage can and cannot achieve. However, these findings may not be true for all cases as some facilities may achieve these tolerance limits. The lack of sufficient available data related to grain and feed handling systems was a limiting factor in this study.
 |
| Figure 2. Output distributions of cumulative probability (%) vs. adventitious presence (%) at four stages in the commodity corn supply chain. Curved lines represent the probability of achieving different AP (%) levels at each stage. Probabilities of maintaining AP below 0.9% and exceeding that limit at each stage are also presented. |
| Table 1. Summary of simulation outcomes: probabilities of achieving 0.9%, 1.5%, 3.0%, and 5.0% tolerance limits at different supply chain stages and configurations, 99% feasible tolerance limits, and outputs of sensitivity analysis. | |||||||
| Supply Chain Stage | Probability to Achieve Tolerance Limits | 99% Feasible Tolerance Limit | Sensitivity Analysis (% Contribution to Total Output Variance) | ||||
| 0.9% | 1.5% | 3.0% | 5.0% | ||||
| Farm | 12.3% | 67.3% | 99% | 99% | 2.25% | Seed Impurity (36.9%) Transportation vehicle (31.6%) Cross-pollination (31.5%) | |
| Grain Elevator | 0.1% | 9.2% | 98.1% | 99% | 3.10% | Seed Impurity (27.1%) Transportation vehicle at grain elevator (23.2%) Transportation vehicle at farm (23.2%) Cross-pollination (23.1%) | |
| Grain Processor | 0% | 0% | 22.1% | 99.7% | 4.75% | Seed Impurity (17.8%) Transportation vehicle at grain elevator (15.8%) Transportation vehicle at grain processor (15.8%) Cross-pollination (15.7%) | |
| Feed Mill | Configuration#1 FarmGrain Elevator Grain Processor Feed Mill | 0% | 0% | 0% | 67.4% | 6.25% | Seed Impurity (13.3%) Transportation vehicle at feed mill (12.3%) Transportation vehicle at grain elevator (12%) Transportation vehicle at farm (12%) |
Configuration#2 FarmGrain Elevator Feed Mill | 0% | 0% | 22.3% | 99.7% | 4.72% | Seed Impurity (18.1%) Transportation vehicle at grain elevator (16%) Transportation vehicle at farm (15.6%) Cross-pollination (15.5%) | |
Configuration#3 Farm Feed Mill | 0% | 0.3% | 76% | 99% | 3.85% | Seed Impurity (22.4%) Cross-pollination (19%) Transportation vehicle at feed mill (18.9%) Transportation vehicle at farm (18.9%) | |
Figure 3 shows the output distribution generated after simulating 50,000 iterations for three supply chain configurations. Findings suggest that, in configuration 1, the probability of achieving 0.9%, 1.5%, and 3.0% tolerance limits in non-GM feed was 0% if the feed mill received ingredients (corn and processed corn products) from the grain processor, grain elevator, and farm. The probability of achieving a 5.0% tolerance limit was 67.4% (configuration 1). In configuration 2, The probabilities of achieving 3.0% and 5.0% tolerance limits were 22.3% and 99%, respectively, if the feed mill received ingredients (corn) from the grain elevator and farm. If the feed mill received corn directly from the farm, the probabilities of achieving 3.0% and 5.0% tolerance limits were 76.1% and 99%, respectively (configuration 3). This finding suggests that bypassing some stages in the supply chain and reducing the number of participants can have a positive impact on the feasibility of achieving 3.0% and 5.0% tolerance limits. A 0.9% tolerance limit was not feasible under any scenario, suggesting that achieving low tolerance limits can increase the risk of non-compliance for non-GM producers and can be challenging in a typical bulk commodity supply chain. The feasibility to achieve a given AP tolerance limit depends on the capability of supply chain participants to manage AP at every stage. The findings also underscore the importance of GM testing every time the product moves from one handler to another to ensure that AP levels fall within desired tolerance limits at the time of receiving.
 |
| Figure 3. Output distributions of cumulative probability (%) vs. adventitious presence (%) for three supply chain configurations. Curved lines represent the probability of achieving different AP (%) levels at each configuration. Probabilities of maintaining AP below 3.0% and exceeding that limit are also presented. |
A sensitivity analysis was conducted to identify the factors that significantly impacted model outputs. In table 1, the ‘sensitivity analysis column’ summarizes the outputs of the sensitivity analysis conducted. The top three to four factors with the highest impact on output variance are listed with their percentage contribution. For example, at the farm stage, 36.9% of output variance was caused by the variance in seed impurity, 31.6% of output variance was caused by the variance in cross-pollination, and 31.5% was due to transportation vehicles. This suggests that seed impurity had the strongest influence on variance in AP levels at the farm stage. At the grain elevator stage, 73.4% of output variance was caused by the interaction of AP-contributing factors at the farm stage such as seed impurity, transportation vehicle, and cross-pollination. Nearly 23% of output variance at the grain elevator stage was attributable to transportation vehicles and 3.4% variance was attributable to unloading and conveying equipment residue. Similarly, at the grain processor stage, 67.2% of output variance was attributable to upstream supply chains factors such as seed impurity, transportation vehicles, and cross-pollination. Similar trends were observed at the feed mill stage, signifying the strong impact of upstream suppliers’ practices and actions on overall AP levels. Among all the upstream supplier factors, seed impurity was the most significant factor. This implied that to attain low AP levels in the finished product at the end of the supply chain, the process must start with very low impurity in the seeds. The most sensitive factors at each stage represent the opportunities for improvement. Focusing on these factors while implementing AP mitigation and segregation strategies can provide better results.
Conclusion
The study offers a comprehensive assessment of complexities surrounding the AP of GM material in non-GM food and feed products within the commodity corn supply chain. Based on the analysis of current practices and factors influencing AP, it is concluded that achieving common trade tolerance limits across the entire supply chain would be challenging. The feasibility of achieving 0.9% and 1.5% was found to be low throughout the supply chain. However, 3.0% and 5.0% tolerance limits were found to be feasible at the farm and grain elevator stages. At the grain processor stage, a 5.0% tolerance limit was found to be feasible; none of the specified tolerance limits were reasonably feasible at the feed mill stage. The 99% feasible tolerance limit was 2.25% at the farm stage, 3.10% at the grain elevator stage, 4.75% at the grain processor stage, and 6.25% at the feed mill stage. More participants in the supply chain add more AP-contributing factors, which limits the feasibility of the supply chain to achieve a specified tolerance limit. One potential approach to overcome this challenge is to bypass certain stages in the supply chain. For instance, if the feed mill received corn directly from the farm, bypassing grain elevator, the probabilities of achieving 3.0% and 5.0% tolerance limits were 76.1% and 99%, respectively. It is crucial to note that the feasibility of achieving a particular AP tolerance limit depends on the ability of each participant in the supply chain to effectively manage AP at every stage. The AP-contributing factors evaluated in this study represent the most significant factors influencing AP levels, but there may be other AP-contributing factors, such as storage and handling actions, which are not specifically evaluated in this study. These factors will add to existing AP in the supply chain and will further reduce the feasibility of achieving specified tolerance limits. The lack of sufficient available data related to equipment residue at the processing and feed mill stage was a limiting factor in gauging the impact of processing equipment on AP levels completely. A more in-depth, case-by-case analysis of AP at the facilities such as grain processors and feed mills will be required to completely understand the feasibility of the supply chain participants to achieve low AP tolerance limits.
References
AFIA. (2022). Trade data: 2022 U.S. animal food industry exports. Retrieved from https://www.afia.org/feedfacts/feed-industry-stats/trade-data/
Boac, J. M., Casada, M. E., Maghirang, R. G., & Harner III, J. P. (2012). 3-D and quasi-2-D discrete element modeling of grain commingling in a bucket elevator boot system. Trans. ASABE, 55(2), 659-672. https://doi.org/10.13031/2013.41367
Bradford, K. J. (2006). Methods to maintain genetic purity of seed stocks. Publication No. 8189. ANR Publication - UC Agriculture & Natural Resources, Agricultural Biotechnology. https://doi.org/10.3733/ucanr.8189
Devos, Y., Demont, M., Dillen, K., Reheul, D., Kaiser, M., & Sanvido, O. (2009). Coexistence of genetically modified (GM) and non-GM crops in the European Union. A review. Agron. Sustain. Dev., 29(1), 11-30. https://doi.org/10.1051/agro:2008051
Devos, Y., Reheul, D., & De Schrijver, A. (2005). The co-existence between transgenic and non-transgenic maize in the European Union: A focus on pollen flow and cross-fertilization. Environ. Biosaf. Res., 4(2), 71-87. https://doi.org/10.1051/ebr:2005013
Dolphin, C. J., Mosher, G. A., Ambrose, R. P., & Ryan, S. J. (2020). Meeting the tolerance: How successful is coexistence in commodity corn handling systems? Appl. Eng. Agric., 36(5), 777-784. https://doi.org/10.13031/aea.14042
Elbehri, A. (2007). The changing face of the U.S. grain system: Differentiation and identity preservation trends. Economic Research Report 35. USDA-ERS. https://doi.org/10.22004/ag.econ.7185
European Commission. (2003). Regulation (EC) No 1829/2003 of the European Parliament and the Council of 22 September 2003 on genetically modified food and feed. Off. J. Eur. Union, L 268/1, Article 1. Retrieved from https://eur-lex.europa.eu/legal-content/EN/TXT/PDF/?uri=CELEX:32003R1829&from=EN
Gupta, P., Hurburgh, C. R., Bowers, E. L., & Mosher, G. A. (2022). Application of fault tree analysis: Failure mode and effect analysis to evaluate critical factors influencing non-GM segregation in the US grain and feed supply chain. Cereal Chem., 99(6), 1394-1413. https://doi.org/10.1002/cche.10601
Hanna, H. M., & Jarboe, D. H. (2011). Effects of full, abbreviated, and no clean-outs on commingled grain during combine harvest. Appl. Eng. Agric., 27(5), 687-695. https://doi.org/10.13031/2013.39566
Harl, N. E., Ginder, R. G., Hurburgh, C. R., & Moline, S. (2001). The StarLinkTM situation. Retrieved from https://www.extension.iastate.edu/grain/files/Migrated/0010star.PDF
Hu, D., Sun, T., Yao, L., Yang, Z., Wang, A., & Ying, Y. (2020). Monte Carlo: A flexible and accurate technique for modeling light transport in food and agricultural products. Trends Food Sci. Technol., 102, 280-290. https://doi.org/10.1016/j.tifs.2020.05.006
Hurburgh Jr., C. R. (2000). The GMO controversy and grain handling for 2000. GMO & Grain 2000. Retrieved from https://www.extension.iastate.edu/grain/files/Migrated/99gmoy2k.pdf
Ingles, M. E., Casada, M. E., & Maghirang, R. G. (2003). Handling effects on commingling and residual grain in an elevator. Trans. ASAE, 46(6), 1625-1631. https://doi.org/10.13031/2013.15625
Ingles, M. E., Casada, M. E., Maghirang, R. G., Herrman, T. J., & Harner Iii, J. P. (2006). Effects of grain-receiving system on commingling in a country elevator. Appl. Eng. Agric., 22(5), 713-721. https://doi.org/10.13031/2013.21986
Kruizinga, A. G., Briggs, D., Crevel, R. W., Knulst, A. C., van den Bosch, L. M., & Houben, G. F. (2008). Probabilistic risk assessment model for allergens in food: Sensitivity analysis of the minimum eliciting dose and food consumption. Food Chem. Toxicol., 46(5), 1437-1443. https://doi.org/10.1016/j.fct.2007.09.109
Laux, C. M., Mosher, G. A., & Hurburgh, C. R. (2015). Application of quality management systems to grain handling: An inventory management case study. Appl. Eng. Agric., 31(2), 313-321. https://doi.org/10.13031/aea.31.10860
Maier, D. E. (2006). Engineering design and operation of equipment to assure grain quality and purity - KP2-2. Proc. 9th Int. Working Conf. on Stored Product Protection. Retrieved from https://citeseerx.ist.psu.edu/viewdoc/download?doi=10.1.1.577.509&rep=rep1&type=pdf
Mosher, G., & Hurburgh, C. (2010). Transgenic plant risk: Coexistence and economy. In D. R. Heldman, & D. G. Hoover (Eds.), Encyclopedia of biotechnology in agriculture and food (1 ed., pp. 639-642). Boca Raton: CRC Press. https://doi.org/10.1081/E-EBAF
OECD. (2021). OECD Seed Schemes 2021: OECD Schemes for the varietal certification or the control of seed moving in international trade. OECD. Retrieved from https://www.oecd.org/agriculture/seeds/documents/oecd-seed-schemes-rules-and-regulations.pdf
Pla, M., Paz, J.-L. L., Peñas, G., García, N., Palaudelmàs, M., Esteve, T.,... Melé, E. (2006). Assessment of real-time PCR based methods for quantification of pollen-mediated gene flow from GM to conventional maize in a field study. Transgenic Res., 15(2), 219-228. https://doi.org/10.1007/s11248-005-4945-x
Syngenta International AG. (2005). Syngenta agrees to settlement with USDA on Bt10 corn. Retrieved from https://www.saveourseeds.org/fileadmin/files/SOS/Dossiers/Mon_863/Syngenta_8April2005.pdf
USDA-AC21. (2005). Global traceability and labeling requirements for agricultural biotechnology-derived products: Impacts and implications for the United States: A report prepared by the USDA Advisory Committee on Biotechnology and 21st Century Agriculture. Retrieved from https://www.usda.gov/sites/default/files/documents/tlpaperv37final.pdf
USDA-AC21. (2016). A framework for local coexistence discussions. A report of the Advisory Committee on Biotechnology and 21st Century Agriculture (AC21) to the Secretary of Agriculture. Retrieved from https://www.usda.gov/sites/default/files/documents/ac21-report-final-local-coexistence.pdf
USDA-AMS. (2018). Final rule: National Bioengineered Food Disclosure Standard. 83 FR 65814, 7 CFR 66. Federal Register: The Daily Journal of the United States Government. Retrieved from https://www.federalregister.gov/documents/2018/12/21/2018-27283/national-bioengineered-food-disclosure-standard
USDA-ERS. (2020). Adoption of genetically engineered crops in the U.S.: Recent trends in GE adoption. Economic Research Service, United States Department of Agriculture. Retrieved from https://www.ers.usda.gov/data-products/adoption-of-genetically-engineered-crops-in-the-us/recent-trends-in-ge-adoption.aspx
USDA-FAS. (2021). 2021: United States agricultural export yearbook. Foreign Agricultural Service, United States Department of Agriculture. Retrieved from https://www.fas.usda.gov/sites/default/files/2022-04/Yearbook-2021-Final.pdf
Vose, D. (2008). Risk analysis: A quantitative guide (3rd ed.). Chichester, England: John Wiley & Sons.
Weber, W. E., Bringezu, T., Broer, I., Eder, J., & Holz, F. (2007). Coexistence between GM and non-GM maize crops – Tested in 2004 at the field scale level (Erprobungsanbau 2004). J. Agron. Crop Sci., 193(2), 79-92. https://doi.org/10.1111/j.1439-037X.2006.00245.x
Wilson, W. W., Janzen, E. L., Dahl, B. L., & Wachenheim, C. J. (2003). Issues in development and adoption of genetically modified (GM) wheats. Agribusiness and Applied Economics Report No. 509. http://www.doi.org/10.22004/ag.econ.23497