A SYSTEMS APPROACH TO MICROBIAL FOOD SAFETY IN PRODUCE: LEVERAGING DATA SCIENCE APPROACHES TO INFORM FOOD SAFETY DECISIONS

• Sponsoring Institution: National Institute of Food and Agriculture
• Project Status: EXTENDED
• Funding Source: OTHER GRANTS
• Reporting Frequency: Annual
• Accession No.: 1020233
• Grant No.: 2019-51181-30016
• Project No.: NYC-143559
• Proposal No.: 2019-03139
• Program Code: SCRI
• Project Start Date: Sep 1, 2019
• Project End Date: Aug 31, 2024
• Grant Year: 2019

Project Director
Wiedmann, M.

Recipient Organization
CORNELL UNIVERSITY
(N/A)
ITHACA,NY 14853

Performing Department
Food Science

Non Technical Summary
While produce represents a healthy food choice, microbial foodborne disease cases and outbreaks are not infrequently linked to contaminated produce. Transmission of microbes causing foodborne illness through produce is a particular risk is produce is often consumed raw; produce, unlike many other foods, thus is often not exposed to a processing step (such as cooking), which reduces the risk of illness due to microbes. Foodborne illness cases and outbreaks negatively affect the US society in many different ways, including (i) the direct negative impact on public health, (ii) the potential to negatively affect the image of produce, therefore reducing produce consumption (and hence the health benefits associated with produce), (iii) negative economic consequences of produce contamination with disease causing bacteria for farmers and produce associated companies, and (iv) even possible negative consequences for export of US produce and other foods.As foodborne pathogens can enter the produce chain at many parts of the food system (such as irrigation water, in the field, in the processing plant, at retail, and even in consumer homes), this project takes a comprehensive systems approach (from field to consumer) to develop and disseminate new tools to reduce the introduction and transmission of microbes causing foodborne illness through different produce types. This main goal will be pursued through (i) research and development of new digital tools to identify and implement effective control strategies that reduce transmission of microbes that cause foodborne illness; (ii) assembly of a library of computational and digital tools applicable to the produce industry, and (iii) training of industry and students in use of digital food safety tools for the produce industry. This project will not only provide US produce farmers and industry with the tools and long term solutions needed to reduce the transmission of microbes causing foodborne illness, but will also train a large number of individuals to develop and apply digital food safety tools to produce. Long-term, this project will (i) reduce foodborne illness cases transmitted through foods and (ii) help the US produce industry to stay competitive as digital tools are being developed and implemented internationally in agriculture and food production.

Animal Health Component

0%

Research Effort Categories

Basic

(N/A)

Applied

100%

Developmental

(N/A)

Classification

Knowledge Area (KA)	Subject of Investigation (SOI)	Field of Science (FOS)	Percent
712	4099	1100	50%
712	1430	2090	30%
712	4099	3020	20%

Knowledge Area
712 - Protect Food from Contamination by Pathogenic Microorganisms, Parasites, and Naturally Occurring Toxins;

Subject Of Investigation
4099 - Microorganisms, general/other; 1430 - Greens and leafy vegetables;

Field Of Science
2090 - Statistics, econometrics, and biometrics; 3020 - Education; 1100 - Bacteriology;

Keywords

food safety

produce

systems approach

Goals / Objectives
Goal 1: Develop a farm-to-consumer hazard transmission model to facilitate a systems approach to identifying environment appropriate, socially acceptable and cost-effective interventions.Goal 2: Improve GIS-based tools and approaches to reduce produce microbial food safety hazards introduction from agricultural water, environmental and animal sources at the pre-harvest level.Goal 3: Develop and validate simulation models for (i) packing houses, (ii) processing plants and (iii) retailers to facilitate use of in silico tools to optimize pathogen sampling plans and control strategies.Goal 4. Assemble a library of existing models and modelling tools applicable to the produce industry, and review these tools with respect to the intended purpose, utility (pros and cons) and current usage, to facilitate computer enabled food safety decision making.Goal 5: Develop a comprehensive on-line and in-person (mixed model) outreach program to provide technical and systems-based produce food safety training and to enable industry to use modeling tools.Goal 6: Develop graduate and undergraduate teaching modules to train students to use and develop computational and modelling tools for produce safety.

Project Methods
This project will be accomplished by developing technical cores (e.g., "modeling" core; "microbiology" core), which will also have the capability to quickly address emerging issues using inter- and transdisciplinary approaches. Goals 1 - 3 will be accomplished through a combination of statistical, mathematical and spatial modeling approaches as well as field studies in different states, across the different compartments of the produce production and processing system; collaboration with the Cornell Digital Ag Initiative will provide an important resource for these efforts. Field sites and specific crop systems will be targeted to (i) lead to findings that can be generalized to multiple crops and to (ii) focus on high priority crops. For Goal 1, we will develop an overall mechanistic understanding of farm-to-consumer hazard transmission that will include economic cost-benefit assessment capabilities and data-supported weighting of interventions based on (i) cost-effectiveness, (ii) sustainability and (iii) social acceptability; this will allow identification of the optimal approach in a given environment/region. We are aware that FDA completes cost-benefit analysis on proposed food safety regulations; our efforts in refining the scientific relationships in this area will allow for more comprehensive economic estimates. Goals 2 and 3 will focus on developing and validating specific models for (i) pre-harvest, including ag water and fields, and (ii) post-harvest, including packing houses, processing plants, and retail. For these two goals, we will use initial models that we have already developed as starting points and will refine and improve these models as well as develop additional new models as needed; targeted field studies and data collection efforts will be conducted to support model development and validation. As part of goals 1 - 3, we will develop, distribute, and implement standards for conduct of field studies to assure that our study as well as future studies can be integrated through meta-analyses to generalize findings and better inform decisions. For goal 4, we will assemble a library of existing models and modelling tools applicable to produce, including appropriate models developed here (e.g., as part of goals 1 - 3); industry surveys will be conducted to assess current usage, needs, and barriers to use of simulation models and other decision-support tools. These data will be used to design the www page and supporting documentation and will help with the design of industry workshops and training programs. Success of the website will be measured by the number of visits to the website and other relevant statistics. Goal 5 will leverage our existing training expertise to develop on-line and in-person programs that will provide systems-based produce food safety training focusing on (i) use, application, and further development of modeling and risk assessment tools to assess and develop produce food safety practices, and (ii) a series of comprehensive training programs on produce food safety systems for growers, packers, processors, and retailers, focusing on science-based strategies, including those derived from our work. Most of these programs will be supplemental to Produce Safety Alliance (PSA) and FSMA training; as appropriate, we will also integrate topics into PSA and FSMA-related training. Published recommendations will be used to develop effective evaluations of (i) knowledge gain, (ii) behavioral change (e.g., implementation of improved food safety practices), and (iii) economic and social impact of the training. Knowledge gain will be assessed through pre- and post-tests administered at every training. A database of people trained will facilitate follow-up assessments for behavioral change; participants will be surveyed 12 months after the training through a on-line survey with questions to assess (i) change of practices (e.g., use of GIS tools); (ii) costs incurred to implement interventions or changes of practice; (iii) quantifiable improvements as a result of change of practice (e.g., fewer audit issues, fewer indicator or pathogen positive results), and (iv) obstacles to behavioral changes and implementation of new practices. Pre-post test comparisons and survey results will be used to identify and implement improvements to all training programs. Data on implementation costs and the associated improvements will be used to conduct a quantitative ex post evaluation of the economic and social impact of the training workshops. The analysis will employ an instrumental variable (IV) method (e.g., Heckman selection model) to allow for endogeneity in program participation and produce unbiased estimates. Goal 6 will develop 4-week teaching modules for undergraduate and graduate students focusing on the basics of risk assessment and modelling with applications to produce safety from farm to table in order to assure a cohort of future food safety professionals that can apply and further develop the tools developed here as well as similar other tools. Modules will be used and tested at the participating universities and made available for free. All modules will undergo peer-review by industry and academic partners included in this project. Anytime modules are used in a class, we will perform (i) student evaluations after completion of the modules and (ii) summative and formative assessment of student learning.

Progress 09/01/22 to 08/31/23

Outputs
Target Audience:The key beneficiaries and target audiences for this project will be fresh produce growers and processors, developers of modeling technologies for use in produce food safety. The project will provide data to help industry to implement sustainable mitigation strategies that are cost-effective, environmentally friendly, and socially acceptable in reducing public health risks, which will benefit the consumers as well. This project also specifically targets produce food safety and quality assurance professionals in industry, academia, and government. Goal 5 focuses on development of comprehensive produce safety extension programs and, in addition, targets retailers and educators. Finally, Goal 6 ("Develop graduate and undergraduate teaching modules to train students to use and develop computational and modelling tools for produce safety") also targets students interested in digital produce food safety. Changes/Problems:Goal 1. We have modified the methodology for the sustainability analysis. We initially considered using a semi-quantitative method, namely expert elicitation, to conduct the sustainability analysis, where the opinions of food industry experts would be quantified. However, we have opted for a fully quantitative, and thus, more rigorous approach. In this new methodology, we are estimating the sustainability impact of each intervention in terms of US dollars based on available literature data. We have identified specific factors that describe the sustainability impacts of each intervention, enabling us to estimate the impacts in US dollars. Any factors that cannot be quantified will be recognized as limitations in our future reports and publications. Goal 2. Nothing to report. Goal 3. Nothing to report. Goal 4. Nothing to report. Goal 5. Nothing to report. Goal 6. Nothing to report. What opportunities for training and professional development has the project provided?Goal 1. Two postdocs, Drs. Ece Bulut and Sarah Murphy have continued training to develop quantitative microbial risk assessment models. A Master of Public Health (MPH) student (Linda Kalunga, Cornell) and a PhD student in Applied Economics and Management (Jieyu Hao, Cornell) received training in sustainability analysis. Goal 2. This project goal has provided training opportunities in GIS and R based modeling tools for a PhD student (Claire Murphy, Virginia Tech) and MS students (Clara Diekman, Camryn Cook, Virginia Tech). Additionally, both students have received basic training in bacterial enumertion and enrichment techniques. Goal 3. Two PhD students (Luke Qian and YeonJin Jung, both at Cornell, are currently being trained on modeling as part of this project component. Goal 4. Two undergraduate students have received training (Sophia Ruser in scoping review and website development and Tyler Wu in using active learning classifiers for screening of articles in literature reviews; both students at Cornell)). Additionally, PhD student (Linda Kalunga, Cornell) received training in economic/sustainability analysis of preharvest produce safety control strategies. Goal 5. A Postdoctoral Research Associate (Dr Magdalena Pajor, Cornell) -, and an Extension Support Specialist (Louise Felker, Cornell), have been actively engaged in the ongoing development and distribution of industry-specific courses. This presents an excellent chance for the Postdoctoral Researcher to not only contribute to the creation of scientific content for professionals beyond academia but also to establish valuable connections in the industry and actively participate in the creation of educational materials. Goal 6. This goal was finalized in the previous reporting period (2021-2022). How have the results been disseminated to communities of interest?For all goals, preliminary results have been shared among the members of the external Advisory Council (comprised of representatives of the produce industry - growers, processors and distributors, trade and research organizations, and government agencies). In addition, a number of preliminary findings have been communicated through abstracts and talks presented at scientific meetings (see below for a complete listing). In 2023, this project also has contributed 5 peer-reviewed publications. In addition, developed videos have been shared on YouTube (combined about 200 views thus far) and with members of the members of the Artificial Intelligence Institute for Next Generation Food Systems. Teaching modules and a virtual library of food safety models and computational decision-support tools, including apps, are shared on the project www site. What do you plan to do during the next reporting period to accomplish the goals?Goal 1. We will complete the manuscripts for the QMRA models and continue to evaluate the sustainability impacts for the identified intervention strategies. Goal 2. Over the next reporting period, we will work to finalize manuscripts associated with Obj 2 outputs including the water quality meta-data method variation analyses (submitted, under review Applied and Environmental Microbiology), and soil surveys (FL submitted Journal of Food Protection, VA accepted, Frontiers). We will also work to publish the meta data associated with our water quality work via a GitHub site (which includes over 100 cleaned and tidied datasets). Goal 3. Finalize manuscripts for publication. Goal 4. We will complete the synthesis of papers on models in food safety identified through a scoping review. We will also complete the website "Virtual Library of Food Safety Models". Goal 5. A dedicated postdoctoral researcher has been hired to this project as previously planned and will be responsible for (i) completing prioritized modules and (ii) beta testing the modules in cooperation with produce safety experts from industry and extension. Through cooperation with the CALS Extension Support developed courses/ teaching modules will be shared on the project www site. Goal 6. This goal was finalized in the previous reporting period (2021-2022).

Impacts
What was accomplished under these goals? Goal 1. We are finalizing two papers on quantitative microbial risk assessments (QMRA) for two produce-pathogen pairs: (i) Escherichia coli O157:H7 and fresh-cut lettuce (EC) and (ii) Listeria monocytogenes and fresh-cut cantaloupe (LM). For the EC QMRA, our findings highlight the significance of initial field contamination, particularly through the irrigation water source (well or surface) and the irrigation system (overhead spray, drip, or furrow), in determining health risks. In the LM QMRA, our model highlights the impact of time and temperature conditions during post-processing stages, including distribution, retail, and home storage. Based on the outcomes of our QMRA models, the project team has identified potential interventions for mitigating health risks and is currently evaluating their sustainability impacts for the fresh-cut lettuce and fresh-cut cantaloupe industry. For fresh-cut lettuce, these interventions involve (i) surface water treatment (using Chlorine, Peracetic acid or UV treatment) for overhead spray irrigation and (ii) transitioning from furrow/overhead spray irrigation to drip irrigation. For fresh-cut cantaloupe, the interventions include (i) moving fresh-cut cantaloupe packages from open refrigeration to closed refrigeration in retail stores and (ii) improving refrigeration during retail (reduced temperature). We are conducting a detailed evaluation of the sustainability implications across three key dimensions: economic, ecological, and social. In each dimension, we are quantifying specific factors (in US dollars) directly affected by each intervention, including private and public costs and benefits for the US fresh-cut lettuce and fresh-cut cantaloupe industry. These factors include the annual cost of water usage (drip irrigation compared to other systems; used in fresh-cut lettuce, intervention (i)) or the annual cost of electricity use (closed versus open refrigeration; used in fresh-cut cantaloupe, intervention (i)), among others. Goal 2. Differences in methodology (e.g., sample volume, detection method) have been shown to affect microbial water quality results, but research is needed to determine how these differences impact the comparability of findings to generate best management practices, and the ability to perform meta-analyses. Here, we address this knowledge gap by compiling and analyzing a dataset representing 2,429,990 unique data points on at least one microbial water quality target (e.g., Salmonella presence, E. coli concentration). Variance partitioning analysis was used to quantify the variance in likelihood of detecting each pathogenic target that was uniquely and jointly attributable to non-methodological versus methodological factors. The strength of the association between microbial water quality and select methodological and non-methodological factors was quantified using conditional forest and regression analysis. Fecal indicator bacteria concentrations were more strongly associated with non-methodological factors than methodological factors. Variance partitioning analysis could not disentangle non-methodological and methodological signals for pathogenic E. coli, Salmonella, and Listeria, suggesting that our current perceptions of foodborne pathogen ecology in water systems are confounded by methodological differences between studies (e.g., only 13% of the total variance in the likelihood of Salmonella detection was uniquely attributed to non-methodological factors), highlighting the need for standardization of methods for microbiological water quality testing for comparison across studies. In this goal, we also evaluated the prevalence of Salmonella and L. monocytogenes (Lm) and the concentration of indicator bacteria (total coliforms, generic E. coli) in agricultural soils, and characterized associations between soil properties and each microbial target in Virginia. Three produce farms, representing different regions and soil types, were sampled four times over one year (October 2021-November 2022). For each individual farm visit, composite soil samples were collected from 20 sample sites per farm per visit for microbial and nutrient analysis (n=240). Samples (25g) were processed for Listeria spp. and Salmonella using a modified FDA BAM method; samples (5g) were enumerated for generic E. coli and total coliforms using Petrifilm. Presumptive Listeria spp. and Salmonella isolates were confirmed by PCR. Soil nutrients from each sample were tested and evaluated for their association with each microbial target. Salmonella, generic E. coli, Listeria spp., and Lm prevalence was 4.2% (10/240), 44.6% (107/240), 10% (24/240) and 2.5% (6/240), respectively. The average concentration of generic E. coli in positive samples was 1.5±0.8 log10 CFU/g. Soil pH was positively associated with Lm [Odds Ratio (OR)=5.5] and generic E. coli (OR=4.9) prevalence; there was no association between Salmonella prevalence and any evaluated factor. Results show that pathogen prevalence was relatively low in unamended soils, and that factors influencing prevalence and concentration varied by microbial target and farm. We also evaluated the prevalence ofSalmonellaand the prevalence and concentration of genericE. coliin Florida's agricultural soils was evaluated to understand the potential risk of microbial contamination at the pre-harvest level. A longitudinal field study was performed in three agricultural areas across Florida. At each location, 20 unique 5x5 m field sampling sites were selected, soil was collected and evaluated forSalmonellapresence (25g) andE. coliand coliform concentrations (5g). Complementary data including weather, adjacent land-use, soil properties, and field management practices were collected from October 2021 to April 2022. The overallSalmonellaand genericE. coliprevalence was 0.4% (1/239) and 11.3% (27/239), respectively; with meanE. coliconcentrations in positive samples of 1.56 log CFU/g. A significant relationship (p<0.05) was observed between genericE. coliand coliforms, and farm and sampling trip. GenericE. coliwas detected in Florida soils throughout the duration of the growing season meaning activities that limit contact between soil and horticultural crops should continue to be emphasized. Goal 3. An agent-based model for the produce section of the retail store has been completed, and the predicted outputs were evaluated using different analyses such as (i) sensitivity analysis, (ii) cluster analysis, and (iii) scenario analysis. Goal 4. To ensure that existing models for inclusion in the virtual library are identified, we first conducted a rapid scoping review of peer-reviewed articles published over the last 10 years by searching the databases PubMed, Web of Science, AGRICOLA, CAB Abstracts, Medline, and FSTA. The search resulted in 6,110 distinct peer-reviewed articles. After title/abstract screening, 215 articles remained. A full-text screening reduced the number of relevant papers to 52. The most common reason for excluding articles was a lack of access to code or a link to the model. For the identified 52 articles, we designed a website to serve as a virtual library of food safety models and computational decision-support tools, including apps. The website is hosted on the www page of the Cornell Institute for Food Safety (https://cals.cornell.edu/institute-for-food-safety). The preliminary virtual library website is organized into multiple tabs based on the microorganism for which the model was intended. Goal 5. The "Introduction to Food Microbiology", "Food Code" and "Additional Strategies to Prevent Contamination" courses have been completed, and await feedback from the Advisory Council (AC). The "Root Cause Analysis" course content has been developed and is approximately 60% complete; it needs to be completed with specific produce specific examples after receiving feedback from the AC. Goal 6. This goal was finalized in the previous reporting period (2021-2022).

Publications

• Type: Journal Articles Status: Published Year Published: 2023 Citation: Cook, C., D.L. Weller, C.M. Murphy, C. Diekman, A.M. Hamilton, M. Ponder, R.R. Boyer, S.L. Rideout, R.O. Maguire, M.D. Danyluk, and L.K. Strawn. 2023. Factors Associated with Foodborne Pathogens and Fecal Indicator Organisms in Virginia Agricultural Soils. Frontiers in Sustainable Food Systems.
• Type: Journal Articles Status: Published Year Published: 2023 Citation: Murphy, C.M., D.L. Weller, R. Ovissipour, R. Boyer, and L.K. Strawn. 2023. Spatial Versus Non-Spatial Variance in Fecal Indicator Bacteria Differs Within and Between Ponds. Journal of Food Protection. 86 (3), 100045.
• Type: Journal Articles Status: Published Year Published: 2023 Citation: Murphy, C.M., Weller, D.L. & Strawn, L.K., 2023. Scale and detection method impacted Salmonella prevalence and diversity in ponds. Science of The Total Environment, p.167812.
• Type: Journal Articles Status: Published Year Published: 2023 Citation: Weller, D.L., Love, T.M., Weller, D.E., Murphy, C.M., & Strawn, L.K. 2023. Scale of analysis drives the observed ratio of spatial to non-spatial variance in microbial water quality: insights from two decades of citizen science data. Journal of Applied Microbiology. 134(10).
• Type: Journal Articles Status: Submitted Year Published: 2024 Citation: Weller, D.L., C.M. Murphy, T.M.T. Love, M.D. Danyluk, and L.K. Strawn. Methodological Differences Between Studies Confound One-Size Fits All Approaches to Managing Surface Waterways for Food and Water Safety. Applied and Environmental Microbiology.
• Type: Journal Articles Status: Submitted Year Published: 2024 Citation: Diekman, C.M., C. Cook, R.W. Worobo, L.K. Strawn, and M.D. Danyluk. Factors Associated with the Prevalence of Salmonella, Generic Escherichia coli, and Coliforms in Floridas Agricultural Soils. Journal of Food Protection.
• Type: Conference Papers and Presentations Status: Published Year Published: 2022 Citation: Ruser, S. Scoping Review to Assess the Ecology of Foodborne Infectious Agents and the State of the Digitally Enabled Food Safety. Ecology & Evolutionary Biology Graduate Student Symposium, Cornell University, Ithaca, NY. December 8 - 9, 2022. Oral presentation.
• Type: Conference Papers and Presentations Status: Published Year Published: 2023 Citation: Ruser, S., Kalunga, L., Grover, L., Ivanek, R. Scoping Review to Assess the State of the Digitally Enabled Food Safety. Sigma Xi Undergraduate Research Symposium, Virtual. April 14-30, 2023. Virtual Slide Show and Personal Video.
• Type: Conference Papers and Presentations Status: Published Year Published: 2023 Citation: 4. Ivanek, R. The Role of Modeling and Risk Assessment in Controlling Outbreaks The National Academies of Sciences, Engineering, and Medicine; Food and Nutrition Board Summer Meeting: Open Session on New and Emerging Technologies in Food Safety, June 23, 2023.
• Type: Conference Papers and Presentations Status: Published Year Published: 2023 Citation: Diekman, C., and M.D. Danyluk. 2023. Investigating the prevalence of Salmonella and E. coli in Floridas soil and identifying key environmental factors. 11th Annual UF/IFAS CREC Research Symposium. P. 19.
• Type: Conference Papers and Presentations Status: Published Year Published: 2023 Citation: Diekman, C., and M.D. Danyluk. 2023. Investigating the prevalence of Salmonella and E. coli in Floridas soil and identifying key environmental factors. 11th Annual UF/IFAS CREC Research Symposium.
• Type: Conference Papers and Presentations Status: Published Year Published: 2023 Citation: Weller, D.L., C.M. Murphy, T. Love, M.D. Danyluk, and L.K. Strawn. 2023. Methodological differences cofound one-size fits all approaches to agricultural water management. International Association of Food Protection Annual Meeting, Toronto, ON. Canada.
• Type: Conference Papers and Presentations Status: Published Year Published: 2023 Citation: Murphy, C. C., Weller, D. L. and Strawn, L. K. 2023. Not All Ponds are Created Equal: Factors Associated with Salmonella Contamination Varies by Pond and Detection Method. International Association for Food Protection 2023 Annual Conference.
• Type: Conference Papers and Presentations Status: Published Year Published: 2023 Citation: Cook, C., Murphy, C.M, Weller, D. L, Ponder, M., Boyer, R. R., Rideout, S., Maguire, R. O., Strawn, L. K. 2023. Soil Nutrient Levels Associated with Salmonella Prevalence and Escherichia coli and Total Coliform Concentrations on Produce Farms. International Association for Food Protection 2023 Annual Conference.
• Type: Conference Papers and Presentations Status: Published Year Published: 2023 Citation: Y. Jung, C. Qian, C. Barnett-Neefs, R. Ivanek, and M. Wiedmann. Validating Agent-Based Model (ABM) that Predicts Listeria spp. Prevalence on Environmental Surfaces in a Retail Store. 2023 International Association for Food Protection Annual Meeting. July 16-19, 2023, Toronto, Ontario, Canada.
• Type: Conference Papers and Presentations Status: Published Year Published: 2023 Citation: Y. Jung, C. Qian, C. Barnett-Neefs, R. Ivanek, and M. Wiedmann. Validating an Agent-Based Model (ABM) that Predicts Listeria spp. Prevalence in a Retail Store. 2023. AI Institute for Next Generation Food Systems open house. Sept 12-13, 2023, Davis, CA (poster).
• Type: Websites Status: Under Review Year Published: 2023 Citation: The website Virtual Library of Food Safety Models, hosted within the www page of the Cornell Institute for Food Safety ( https://cals.cornell.edu/institute-for-food-safety , is currently open for internal use by the project team)

Progress 09/01/21 to 08/31/22

Outputs
Target Audience:The key beneficiaries and target audiences for this project are fresh produce growers and processors, and developers of modeling technologies for use in produce food safety. The project will provide data to help industry to implement sustainable mitigation strategies that are cost-effective, environmentally friendly, and socially acceptable in reducing public health risks, which will benefit the consumers as well. This project also specifically targets produce food safety and quality assurance professionals in industry, academia, and government. Goal 5, which focuses on development of comprehensive produce safety extension programs in addition targets retailers and educators. Finally, Goal 6 ("Develop graduate and undergraduate teaching modules to train students to use and develop computational and modelling tools for produce safety") also targets students interested in digital produce food safety. Changes/Problems:For Goal 2, the number of Salmonella and Listeria identified in soils was significantly lower than anticipated. This will limit the amount of model work that can be completed. While no major overall changes have been made, initial project delays due to COVID have carried over into subsequent years, which will mean that we will likely need to request a no-cost extension. What opportunities for training and professional development has the project provided?Goal 1. Two postdocs, Drs. Ece Bulut and Sarah I. Murphy have continued training to develop quantitative microbial risk assessment models. Master of Public Health (MPH) student (Linda Kalunga) received training in agent-based modeling applied to food safety. ?Goal 2. This project goal has provided training opportunities in GIS and R based modeling tools for a PhD student (C. Murphy) and MS students (C. Diekman, C. Cook) who were supported on this grant. A. Hamilton, a postdoc, assisted in data cleaning, and was provided training on tidy data, and data structure. Additionally, both students have received basic training in bacterial enumeration and enrichment techniques. Goal 3. This goal has trained two researchers (Genevieve Sullivan and Cecil Barnett-Neefs), which have both graduated (with a PhD and MS, respectively); two more PhD students (Luke Qian and YeonJin Jung are currently being trained on modeling as part of this project component. Goal 4. An undergraduate student (Sophia J. Ruser) and a high school student (Lily Grover) have been trained in food safety and in conducting a scoping review. Goal 5. Two graduate students, Samantha Bolten and Jonathan Sogin, continued training to develop courses for industry personnel and continued to develop connections with industry leaders available to them via the Advisory Council. Goal 6. Two PhD students at Cornell (Luke Qian and Samantha Bolten) were actively involved in module development. In addition, a researcher (Cecil Barnett-Neefs) has been trained in the development of video-based educational materials. Finally, one student at Virginia Tech (C. Murphy) helped in one of the undergraduate-level produce food safety module discussion sessions. How have the results been disseminated to communities of interest?For all goals, preliminary results have been shared among the members of the external Advisory Council (comprised of representatives of the produce industry - growers, processors and distributors, trade and research organizations, and government agencies). In addition, a number of preliminary findings have been communicated through abstracts and talk presented at scientific meetings (see below for a complete listing). In 2022, this project also has contributed four peer-reviewed publications. In addition, developed videos have been shared on YouTube (combined about 100 views thus far) and with members of the members of the Artificial Intelligence Institute for Next Generation Food Systems and teaching modules are shared on the project www site. What do you plan to do during the next reporting period to accomplish the goals?Goal 1. Over the next reporting period, we will finalize manuscripts for the developed QMRA models for EC and LM. Next, we will use predictions from the QMRA models about the risk reduction associated with certain mitigation strategies, as well as information from literature and expert opinion elicitation, to evaluate the indicators of food safety sustainability (social, economic, and environmental) for those mitigation strategies. Goal 2. Over the next reporting period, we will submit manuscripts on the water quality meta-data effort for each microbial target, and method variation analyses for water, and for soil surveys in VA and FL. We will also develop predictive models with different levels of complexity (allowing for use by stakeholders with different levels of expertise); these efforts will initially focus on Listeria predictive models (as methodological signals are not confounding environmental/other factors). Goal 3. We plan to complete the Agent-based model for produce section of retail environments and evaluate it for prediction of the effectiveness of different Listeria control strategies. ?Goal 4. We will complete the scoping review of digitally enabled models and modeling tools. We also expect to create and apply a web-based survey to assess produce industry practices regarding the use of modeling tools and barriers to using them. Goal 5. We plan to hire a postdoctoral researcher that will be dedicated to this project, the modules will be completed and beta tested with produce safety experts from industry and extension. Goal 6. We plan to complete the graduate students modules for produce safety and digital produce safety and subsequently launch a communication campaign to encourage adoption and use of the teaching modules developed as part of this project.

Impacts
What was accomplished under these goals? Goal 1.We are finalizing papers describing quantitative microbial risk assessments (QMRA) for two produce-pathogen pairs: (i) Escherichia coli O157:H7 and fresh-cut lettuce (EC) and (ii) Listeria monocytogenes and fresh-cut cantaloupe (LM). Both models are aimed to evaluate the human health risk from consumption of contaminated lettuce/cantaloupe. QMRA for EC: We revised the pre-harvest module of the QMRA to better represent the field contamination of lettuce via four contamination sources (i.e., contamination via irrigation water, animal feces from field application of manure, wildlife intrusion and runoff from adjacent cattle farm). This included changes in the conceptual model, repurposing parameters/variables used in the model and revalidating the pre-harvest module. In addition, we included parameters and variables to develop alternative scenarios to the baseline model, including variables/parameters related to furrow and drip irrigation and cattle vaccination for pre-harvest module, as well as chlorine wash alternatives and consumer wash in post-harvest module. Preliminary results indicate the importance of the irrigation system used (spray, drip or furrow) in determining the food safety risk associated with consumption of lettuce. QMRA for LM: We expanded the QMRA model to include (i) option of cantaloupe following either a scheme of field to packinghouse to fresh-cut facility OR field to cooling facility to fresh-cut facility, (ii) option of equipment cleaning and/or sanitation at the packinghouse and fresh-cut facility, and (iii) cross-contamination at the packinghouse and fresh-cut facility. We extracted data from the literature for model validation and prepared a set of scenarios for assessment based on input from the project advisory council as well as literature review. Preliminary sensitivity analysis identified that time and temperature conditions during distribution, retail, and home storage, and the initial number of LM cells on the total cantaloupe surface area in a bin had the greatest impacts on LM per contaminated fresh-cut cantaloupe serving. Scenario analysis showed that the relative degree of clustering affects the risk of listeriosis per serving and the predicted number of illnesses and deaths. Assessment of interventions demonstrated that reducing temperature and/or time conditions post-processing can be an effective strategy for reducing LM risk. Goal 2 includes efforts related to both water and soil related pathogen sources. For water related pathogen sources, a 2,429,990 sample dataset representing >500 studies was compiled; each sample had data on at least one microbial water quality (MWQ) target (e.g., Salmonella presence, fecal indicator bacteria [FIB] levels). Variance partitioning analysis showed that environmental and methodological signals could not be disentangled for E. coli (EC), Salmonella, and Listeria, limiting our current understanding of and ability to base guidance on foodborne pathogen ecology in freshwater systems. For example, 18% of variance in Salmonella detection was jointly attributable to methodological and non-methodological factors. Despite this, opportunities for site or waterway-specific management exist because substantial variance in MWQ was still uniquely attributable to environmental factors (e.g., 13% for Salmonella). In contrast, FIB levels were more strongly associated with environmental, as opposed to methodological factors by condition forest. Our findings suggest that metadata collection and management should be standardized across studies (min set of attributes), and comparability data is needed to develop water quality risk assessments. With regard to soil related pathogen sources, the prevalence of EC,Salmonella and Listeria spp., and the concentration of EC and total coliform bacteria in Florida and Virginia agricultural soils was evaluated to understand the risk of microbial contamination at the pre-harvest level. A longitudinal field study was performed in each state consisting of three geographically distributed agricultural areas across each state. At each location, 20 unique 5x5 m field blocks were selected, soil was collected and evaluated for Salmonella and Listeria spp. presence (25g) and ECand total coliform bacteria enumeration (5g). Complementary data included weather, adjacent land-use, soil micro-nutrients, and field management practices from October 2021 to April 2022 in Florida and October 2021 to October 2022 in Virginia. Analyses are ongoing in Virginia, as sampling just concluded. In Florida, the overall prevalence of Salmonella and EC were 0.418% (1/239) and 11.3% (27/239), respectively; mean EC population levels in positive samples were 1.56 log CFU/g. Investigation of the relationship between farm and sampling trip showed that both farm and sampling trip were significant (p<0.05) for ECand coliforms and between farm and sampling trip significant (p<0.05) for coliforms only. For example, in Florida, variation in the prevalence of ECand changes in coliform populations between farms suggests the factors influencing the soil environment at the three farms are different. While Salmonella was only identified once, ECwas detected in Florida soils throughout the duration of the growing season meaning activities that limit contact between soil and crops should continue to be emphasized. Goal 3.We published two papers in 2022, which detail development and use of Agent-Based models (ABMs) to characterize Listeria transmission and control strategies in packing houses. In addition, we have developed an agent-based model on Listeria transmission in retail environments; we currently finalizing the parameter estimates for this model. Initial model runs suggested that limiting the contamination from the incoming supply and improving the cleaning schedule can be effective in reducing Listeria contamination in retail environments. Goal 4.We are conducting a scoping review to identify, summarize and assess digitally enabled models and modeling tools applicable to food safety. We developed a literature search strategy to identify the peer-reviewed articles describing food safety modeling tools by searching across six databases (PubMed, Medline, Web of Science, FSTA, CAB Abstracts and AGRICOLA). To be eligible, a study needs to be published in a peer-reviewed journal within the last 10 years, describe primary research, be written in English, and be about reusable modeling tools developed for various foodborne pathogens or viruses found in food/beverage or agricultural matrices. We first conducted a pilot scoping review that identified a total of 687 articles, screening identified 34 studies eligible for data extraction. The results of the pilot in terms of the identified and missed studies were used to improve the search algorithm and re-run the search. This resulted in the identification of 6,110 studies for screening. A total of 2,358 duplicates were removed, and the remaining articles are currently being screened by title and abstract against the eligibility criteria. Goal 5. The prerequisite courses (Introduction to Food Microbiology, Systems Approach to Produce Safety, Produce Traceability, and Root Cause Analysis Methods) have been prioritized for completion.Introduction to Food Microbiology content is complete but must be finalized with produce specific visual aids and produce specific examples.Produce Traceability and Systems Approach to Produce Safety are approximately 50% of the content developed.Feedback from the Advisory Council was incorporated into the materials that were generated up to this point. Goal 6.We have modules on produce safety (3 lectures) and digital produce safety (2 lectures) on our project www site (https://cals.cornell.edu/produce-safety-coe/training/teaching-modules). We have also developed a lecture and module on food systems digital twins for graduate and undergraduate students as well as industry, which have been posted on YouTube.

Publications

• Type: Journal Articles Status: Published Year Published: 2022 Citation: Barnett-Neefs, C.W., Sullivan, G., Zoellner, C., Wiedmann, M., Ivanek, R. Using Agent-Based Modeling to compare corrective actions for Listeria contamination in produce packinghouses. PLOS ONE, 2022. https://doi.org/10.1371/journal.pone.0265251
• Type: Journal Articles Status: Published Year Published: 2022 Citation: Barnett-Neefs, C., Wiedmann, M., Ivanek, R. Examining Patterns of Persistent Listeria Contamination in Packinghouses using Agent-Based Models. Journal of Food Protection 2022 Dec 1;85(12):1824-1841, doi: 10.4315/JFP-22-119. https://pubmed.ncbi.nlm.nih.gov/36041081/
• Type: Conference Papers and Presentations Status: Published Year Published: 2022 Citation: Ivanek R., Acharya J., Wiedmann M., Zhao Q., Zohdi T., Nitin N., Adalja A., Smith A., Kalunga L. Digital twin models and integrated AI for food safety across the supply chain. AIFS site visit, UC Davis, Sep 20, 2022. Poster presentation
• Type: Conference Papers and Presentations Status: Published Year Published: 2022 Citation: Ivanek, R. In Silico Models for Design and Optimization of Science-Based Listeria Environmental Monitoring Programs in Fresh-Cut Produce Facilities. Invited talk at ASM Microbe 2022, June 11, 2022
• Type: Conference Papers and Presentations Status: Published Year Published: 2021 Citation: Bulut E, Murphy S.I., Strawn L.K., Danyluk M., Wiedmann M. and Ivanek R. Identifying preharvest risk reduction strategies for Escherichia coli O157:H7 on fresh-cut lettuce in the US industry through application of QMRA. Society for Risk Analysis (SRA) 2021 Annual Meeting, Dec 5-9, 2021, Virtual:SRA.org. Poster presentation
• Type: Conference Papers and Presentations Status: Published Year Published: 2021 Citation: Murphy S.I., Bulut E., Strawn L.K., Danyluk M., Wiedmann M. and Ivanek R. Utilizing QMRA to identify postharvest risk reduction strategies for Listeria monocytogenes on fresh-cut cantaloupe in the US industry. Society for Risk Analysis (SRA) 2021 Annual Meeting, Dec 5-9, 2021, Virtual:SRA.org. Poster presentation
• Type: Conference Papers and Presentations Status: Published Year Published: 2022 Citation: Ruser, S. Scoping Review to Assess the Ecology of Foodborne Infectious Agents and the State of the Digitally Enabled Food Safety. Ecology & Evolutionary Biology Graduate Student Symposium, Cornell University, Ithaca, NY. Oral presentation.
• Type: Conference Papers and Presentations Status: Published Year Published: 2022 Citation: Diekman, C., and M.D. Danyluk. 2022. Investigating the Prevalence of Salmonella and E. coli in Floridas Soil and Identifying Key Environmental Factors. International Association for Food Protection Annual Meeting, Pittsburgh, PA. P1-131
• Type: Conference Papers and Presentations Status: Published Year Published: 2022 Citation: Cook, C., M.A. Ponder, C.M. Murphy, A. Hamilton, R.R. Boyer, S. Rideout, R. Maguire, and L.K. Strawn. 2022. Associations between Soil Nutrient Levels with Escherichia coli and Total Coliform Concentrations and Listeria and Salmonella Prevalence. International Association for Food Protection Annual Meeting, Pittsburgh, PA. P1-61
• Type: Conference Papers and Presentations Status: Published Year Published: 2022 Citation: Murphy, C., D. Weller, L.K. Strawn. 2022. Spatial Versus Non-Spatial Variance in E. coli Levels Differs by Scale of Analysis in Virginia Ponds. International Association for Food Protection Annual Meeting, Pittsburgh, PA. P1-63.
• Type: Conference Papers and Presentations Status: Published Year Published: 2022 Citation: Jung, Y., Qian, C., Barnett-Neefs, C., and Wiedmann, M. 2022. Developing an Agent-Based Model to Assess Listeria Control Strategies in Retail Stores. International Association for Food Protection Annual Meeting, Pittsburgh, PA. P3-136
• Type: Conference Papers and Presentations Status: Published Year Published: 2022 Citation: Weller, D.W., C. Murphy, D. Danyluk, L.K. Strawn. 2022. Agricultural Water Quality for Produce: Recent Advances, Current Challenges, and Future Opportunities. International Association for Food Protection Annual Meeting, Pittsburgh, PA. S29.

Progress 09/01/20 to 08/31/21

Outputs
Target Audience:Goal 1. The key beneficiaries will be fresh produce growers and processors, who will be better able to mitigate E. coli O157:H7 or L. monocytogenes contamination of fresh produce. This project specifically provide data for identification of potential contamination pathways for pathogens entering farm-to-fork food production chain. The project also provide data to help industry to implement sustainable mitigation strategies that are cost-effective, environmentally friendly and socially acceptable in reducing public health risks. Thus, the sustainable and safe produce supply will benefit the consumers as well as the society as a whole. Goal 2. This goal will directly benefit produce growers by reducing the risk of field contamination by pathogens through agriculture water. In addition, this goal will also indirectly benefit processing facilities as fewer pathogens will be introduced into these facilities through the raw material, and consumers, who will benefit from having safer produce. Goal 3. Target audiences will be processors, packing houses, and retail food safety managers; these groups are intended users of the agent-based models developed as part of this goal. Goal 4. Target audiences will be produce food safety and QA professionals in industry, academia, and government. Goal 5. No change from the prior reporting period. The beneficiaries of the extension and outreach programs are stakeholder groups across the produce production, processing, and distribution continuum, including (i) growers, (ii) processors and packing house operators, (iii) retailers, and (iv) educators. Students that participate in the project are additional beneficiaries. Goal 6. Academic educators, undergraduate and graduate students focusing on produce food safety will directly benefit from the products of this goal as we will develop modules that educators can use to teach undergraduate and graduate students advanced topics related to produce safety and statistical modelling in the context of produce safety. Indirectly, the fresh produce industry will also benefit as they will be able to hire new personnel that are better trained to tackle real world issues in produce safety and using modeling tools to enhance produce safety. Changes/Problems: Nothing Reported What opportunities for training and professional development has the project provided?Goal 1. Two postdocs, Drs. Ece Bulut and Sarah Murphy have been trained in setup of a framework for sustainability assessment, framing of the risk questions and the development of farm-to-fork conceptual models. Additionally, an undergraduate student (Hannah Rosenthal) has been trained in produce food safety research, focusing on the consumer perceptions, beliefs and behaviors. Goal 2. This project goal has provided training opportunities in GIS and R based modeling tools for a PhD student (C. Murphy) and MS student (C. Diekman) who were supported on this grant. Additionally, both students have received basic training in bacterial enumeration and enrichment techniques. C. Murphy helped in one of the undergraduate-level produce food safety module discussion sessions. Goal 3. As part of this goal, we trained 3 graduate students as well as one undergraduate student (all at Cornell) in development of agent-based models. Goal 4. Postdoc, Dr. Sarah Murphy and PhD student, Luke Qian, have received training in features of different types of models and the basics of web design. Goal 5. Three graduate students, Samantha Bolten, Yadwinder Singh Rana, and Jonathan Sogin, continued training to develop courses for industry personnel and continued to develop connections with industry leaders available to them via the Advisory Council. Goal 6. One graduate student, Samantha Bolten, was trained in developing and finalizing the three lectures in the undergraduate-level produce safety module. Several graduate students participated in the discussion-like recordings of these three lectures, including: Jonathan Sogin, Yadwinder Singh Rana, Samantha Bolten, Alexandra Belias, Luke Qian, and Claire Murphy. One postdoctoral fellow, Sophia Harrand, and one graduate student, Samantha Bolten, were responsible for creating the student knowledge assessments and the module evaluation survey for the undergraduate-level produce safety module. How have the results been disseminated to communities of interest?Overall, results to date have been communicated with the project Advisory Council, which includes 20 members; so far this group has met 3 times. Details on specific communication that occurred for the different goals are provided below. Goal 1. Preliminary results have been shared among the members of the external Advisory Council (comprised of representatives of the produce industry - growers, processors and distributors, trade and research organizations, and government agencies). Additionally, preliminary results have been shared among researchers and students in the academic circles and at a conference (Annual Meeting of the International Association of Food Protection, 2021). Goal 2. Preliminary results have been communicated to the Advisory Council, and project collaborators. Two publications were submitted and accepted in 2021 based on project efforts. Future communications and presentations are planned for International Association of Food Protection (IAFP) 2022, including submission of preliminary findings from the soil/micronutrient sampling effort. Goal 3. Results have been shared through a peer-reviewed manuscript: Sullivan G, Zoellner C, Wiedmann M, Ivanek R. 2021. In silico models for design and optimization of science-based Listeria environmental monitoring programs in freshcut produce facilities. Appl Environ Microbiol 87:e00799-21. https://doi.org/10.1128/AEM.00799-21. Goal 4. Nothing to report. Goal 5. Learning objectives and course outcomes have been communicated to the Advisory Council and project collaborators. Preliminary courses have been distributed to specific members of the Advisory Council, project collaborators, and other industry personnel. Goal 6. All virtual lectures in the undergraduate-level produce safety module were distributed to the Advisory Council to obtain feedback, and this feedback has been logged and will be used to aid in the development of future modules. The undergraduate-level produce safety module (https://foodsafety.foodscience.cornell.edu/produce-safety-coe/teaching-modules/) was provided to summer graduate student interns at iFood Decisions and piloted in an experiential learning program at Virginia Polytechnic Institute and State University in June-August of 2021. Lectures were also made available through YouTube (https://www.youtube.com/channel/UCCmWOQGV7hh6OwjoDJ4GtDw/videos) on the Produce Safety Center of Excellence channel. What do you plan to do during the next reporting period to accomplish the goals?Goal 1. Over the next reporting period, we will complete the quantitative microbial risk assessment (QMRA) models for the E. coli O157:H7-lettuce and L. monocytogenes-cantaloupe commodity-pathogen pairs to identify promising mitigation strategies that reduce the human health risk and start evaluating the indicators of food safety sustainability (social, economic and environmental) for those mitigation strategies. Goal 2. Over the next reporting period, we will continue our analyses and draft manuscripts of the water quality meta-data effort for each microbial target, and method variation analyses. We will also continue to develop models based on different levels of complexity to the produce industry; including, first level: temperature, (air/water), rainfall, other characteristics (e.g., water type); second level: first level plus physiochemical parameters (e.g., dissolved oxygen); third level: first and second level plus microbial test data; four level: first, second and third levels plus land-use data. These models range in the levels of difficulty for the stakeholders, for example, it is fairly easy to obtain weather data from public weather stations (first level), to increasing difficulty to monitor and obtain land-use data from GIS software. We will also continue the soil sampling/micronutrients efforts. Additionally, progress has been made on our goal of investigating pathogen/indicator organism associations with soil micro-nutrients. Sampling efforts are underway on three produce farms in each state, Florida and Virginia. Within each produce farm, twenty 5 m by 5 m plots are randomly selected and five soil samples collected per plot (n=100 soil samples). Soil will be tested using standard methods for E. coli and total coliform concentration, and Salmonella and Listeria presence (only Salmonella in FL); as well as, >20 different micronutrients (similar to Liao et al., 2021; Nature Microbiology). The study will run for one year with longitudinal sampling performed during the produce production season at four different times. Goal 3. We will continue with development of an ABM for retail produce sections in order to model Listeria transmission and develop optimal Listeria sampling strategies and intervention plans. Goal 4. We will start collating existing models and modelling tools in the area of food safety. This may involve some form of a systematic/scoping literature review but we also anticipate that literature identified in Goal 1 will be valuable for Goal 4 as well. We also expect to create and apply a web-based survey to assess produce industry practices regarding the use of modeling tools and barriers to using them. Goal 5. Over the next reporting period, we will finalize the courses that have been started already and will begin to develop those that have not been developed yet. The basic introductory produce safety courses that all participants take will be prioritized for completion first. The subsequent specialized courses for the 4 tracks will be developed next, and finally the advanced courses. The courses will be shared with members of the advisory council, project collaborators and small test groups of industry personnel to solicit feedback. The finalized courses will be made available to the public through the established Produce Safety Center of Excellence https://foodsafety.foodscience.cornell.edu/produce-safety-coe/. Goal 6. During the next reporting period, we will finalize the graduate-level produce safety module, and plan to pilot this module in academic institutions as we did with the undergraduate-level produce safety module. In addition, we expect to have the undergraduate-level statistical modeling module finalized by the next reporting period.

Impacts
What was accomplished under these goals? Goal 1. We have continued the work on the quantitative microbial risk assessments (QMRA) for two produce-pathogen pairs: (i) Escherichia coli O157:H7 and fresh-cut lettuce (EC) and (ii) Listeria monocytogenes and fresh-cut cantaloupe (LM). The exposure assessment evaluated the distribution of the pathogens along the whole supply chain, from pre-harvest to storage at home. Then, dose response models were applied to evaluate the human health risk from consumption of contaminated lettuce/cantaloupe. The models were developed in @Risk software and were parameterized based on published literature and expert opinion. QMRA for EC: We evaluated introduction of contamination via animal feces (from field application of manure, wildlife intrusion, and runoff from adjacent cattle farms) and irrigation at the pre-harvest. Accordingly, the model evaluated several pre-harvest interventions: (i) growing the produce in non-amended soils, (ii) eliminating runoff from adjacent dairy farms, (iii) keeping wildlife off farms and (iv) cattle vaccination. EC concentration in a single serving of fresh-cut lettuce (100 grams) was used to evaluate the impact of each intervention on the risk of human illness in the US annually. Preventing wildlife intrusion, vaccinating cattle and eliminating runoff were identified as the most effective strategies for reducing human illness cases. QMRA for LM: This model accounts for the effects of conditions during transportation and storage as well as fresh-cut preparation (chlorine washing and cutting). The probability of listeriosis was estimated from consumption of a LM-contaminated fresh-cut cantaloupe serving (134 g). The preliminary QMRA model estimated a median of 3.03 log10 CFU of LM (5th, 95th, 99th percentiles: 0.90, 7.18, 9.91) per contaminated serving of fresh-cut cantaloupe at consumption. Notably, the probability of ≥1 death annually attributed to fresh-cut cantaloupe was 12.1%. Sensitivity analysis identified that time and temperature after packaging (retail, distribution, home storage), and the initial concentration of LM on whole cantaloupes at point-of-harvest had the greatest impacts on occurrence of ≥1 death annually attributed to fresh-cut cantaloupe. Goal 2. We conducted a search of available datasets (both published and un-published) on the concentration of indicator organisms and presence of pathogens in surface water surfaces in the United States, and North America, to characterize and identify spatio-temporal factors (e.g., soil properties, meteorological events, and adjacent land-use), and build models to predict pathogen presence (e.g., Listeria monocytogenes, Salmonella), and generic Escherichia coli levels. Raw data was collected from 96 candidate datasets and represented all US states, 3 Canadian provinces, and an additional dataset from Mexico yielding approximately 300,000 individual data points. All data was organized in Excel for importation into the statistical database R (to allow ease of storage, analysis and visualization). Additionally, R code was developed to mine adjacent land-use buffered at 122, 366, and 1098 meters from the sampled GPS location using the National Land Cover Database (NLCD). This was exacted for all data with GPS points (approximately 1/3). Historical weather data was also pulled from the nearest weather station using the date the sample data point was collected. Two publications have results from efforts. Goal 3. For this goal, we have completed Agent-based models (ABMs) for fresh cut facilities, which has been published. More specifically, two ABMs of fresh-cut produce facilities were developed as a way to simulate the dynamics of Listeria by modeling the different surfaces of equipment and employees in a facility as agents. We also developed ABMs of Listeria contamination dynamics for two produce packinghouse facilities; models were parameterized using observations, values from published literature and expert opinion. Once validated with historical data from Listeria environmental sampling, each model's baseline conditions were investigated and used to determine the effectiveness of corrective actions in reducing prevalence of agents contaminated with Listeria and concentration of Listeria on contaminated agents. Corrective actions targeting Listeria introduced in the facility with raw materials, implementing risk-based cleaning and sanitation, and modifying equipment connectivity were shown to be most effective in reducing Listeria contamination prevalence. Overall, our results suggest that a well-designed cleaning and sanitation schedule, coupled with good manufacturing practices can be effective in controlling contamination, even if incoming Listeria spp. on raw materials cannot be reduced. We also initiated development of an ABM for retail produce sections in order to model Listeria transmission and develop optimal Listeria sampling strategies and intervention plans. Goal 4. Initial steps were made on (i) preparation of the library of models and modelling tools and (ii) design of the model library website. Briefly, a preliminary list of existing models and an initial design for the model library website were prepared. Feedback was obtained from the project Advisory Council. Goal 5. Work continued on the development of a national produce safety certificate program that provides training and certification to individuals associated with all segments of the produce industry. The previous reporting period established a preliminary curriculum consisting of 4 specialized certificate tracks; these tracks remain unchanged. After several discussions with the Advisory Council and project collaborators, changes to the preliminary curriculum were made. An additional course, Root Cause Analysis Methods, was added to the basic produce safety courses. The Produce Safety Alliance Grower Training was added as an alternative to GAPs training for the produce grower tract. Food Code training was combined with Additional Strategies to Prevent Contamination for the produce retailer tract. Two additional courses, Product Sampling Strategies, and Food Safety Communication and Enterprise Risk Managements, were added to the Advanced course offerings. In addition to curriculum development, learning objectives and course outcomes were established for each of the courses in the curriculum. Significant progress on the development of four courses, Systems Approach to Produce Safety, Produce Traceability, Food Code and Additional Strategies to Prevent Contamination (retail tract), and Environmental Monitoring, was made. Feedback from individuals of the Advisory Council and project collaborators was solicited and integrated for these preliminary-stage courses. Goal 6. Progress was made on development of the produce safety and modeling modules proposed under goal 6. First, the undergraduate-level produce safety module (which includes three ~45-minute lectures recorded using a discussion-like format) has been completed. The undergraduate-level produce safety module was piloted in an experiential learning program at Virginia Polytechnic Institute and State University in June-August of 2021. For knowledge assessment of learning outcomes for the undergraduate-level produce safety module, pre- and post-assessments were created and provided to students who participated in the experiential learning program. We also obtained student feedback by having students fill out evaluation surveys after completion of the module. The surveys indicated that students enjoyed the use of real-life examples of outbreaks to help teach principles of produce safety and outbreak investigation. In addition, learning outcomes were defined for the undergraduate-level and graduate-level modules focused on teaching principles of modeling in the context of produce safety. The first framework of the undergraduate-level modeling module is currently in preparation.

Publications

• Type: Journal Articles Status: Accepted Year Published: 2021 Citation: D.L. Weller, C. Murphy, S. Johnson, H. Green, E. Michalenko, T. Love, and L.K. Strawn. 2021. Land-use, Weather, and Water Quality Factors are Associated with Fecal Contamination of Northeastern Streams that Span an Urban-Rural Gradient. Frontiers in Water. Accepted November 15, 2021.
• Type: Journal Articles Status: Published Year Published: 2021 Citation: Sullivan G, Zoellner C, Wiedmann M, Ivanek R. 2021. In silico models for design and optimization of science-based Listeria environmental monitoring programs in freshcut produce facilities. Appl Environ Microbiol 87:e00799-21. https://doi.org/10.1128/AEM.00799-21.
• Type: Websites Status: Published Year Published: 2021 Citation: The model algorithm (code) for the packing house agent-based models was written in the open source program NetLogo 6.1.1 and is available on GitHub (https://github.com/IvanekLab/CPS_2019_OpenData).
• Type: Websites Status: Published Year Published: 2021 Citation: Undergraduate-level produce food safety module: https://foodsafety.foodscience.cornell.edu/produce-safety-coe/teaching-modules/
• Type: Conference Papers and Presentations Status: Other Year Published: 2021 Citation: Oral presentation by Dr. Bulut: E Bulut, S I Murphy, L K Strawn, M Danyluk, M Wiedmann and R Ivanek. A farm-to-fork quantitative microbial risk assessment model for Escherichia coli O157:H7 on fresh-cut lettuce. International Association for Food Protection (IAFP) Annual Meeting, Jul 18-21, 2021, Phoenix AZ.
• Type: Conference Papers and Presentations Status: Other Year Published: 2021 Citation: Oral presentation by Dr. Murphy: S I Murphy, E Bulut, L K Strawn, M Danyluk, M Wiedmann and R Ivanek. Farm-to-consumer quantitative risk assessment model for Listeria monocytogenes on fresh-cut cantaloupe. International Association for Food Protection (IAFP) Annual Meeting, Jul 18-21, 2021, Phoenix AZ.
• Type: Journal Articles Status: Accepted Year Published: 2021 Citation: Murphy, C.M., L.K. Strawn, T.K. Chapin, R. McEgan, S. Reddy, L. Friedrich, L. Goodridge, D.L. Weller, K. Schneider, and M.D. Danyluk. 2021. Factors associated with E. coli levels in and Salmonella contamination of agricultural water differed between North and South Florida waterways. Frontiers in Water. Accepted November 16, 2021.

Progress 09/01/19 to 08/31/20

Outputs
Target Audience:Goal 1. The key beneficiaries will be fresh produce growers and processors, who will be better able to mitigate E. coli O157:H7 or L. monocytogenes contamination of fresh produce. This project specifically provide data for identification of potential contamination pathways for pathogens entering farm-to-fork food production chain. The project also provide data to help industry to implement sustainable mitigation strategies that are cost-effective, environmentally friendly and socially acceptable in reducing public health risks. Thus, the sustainable and safe produce supply will benefit the consumers as well as the society as a whole. Goal 2. This goal will directly benefit produce growers by reducing the risk of field contamination by pathogens through agriculture water. In addition, this goal will also indirectly benefit processing facilities as fewer pathogens will be introduced into these facilities through the raw material, and consumers, who will benefit from having safer produce. Goal 3. Nothing to report. Goal 4. Nothing to report. Goal 5. The beneficiaries of the extension and outreach programs are stakeholder groups across the produce production, processing, and distribution continuum, including (i) growers, (ii) processors and packing house operators, and (iii) retailers as well as consumer outreach efforts that will also include students that participate in the project. Goal 6. Academic educators, undergraduate and graduate students will directly benefit from the products of this goal as we will develop modules that educators can use to teach undergraduate and graduate students on advanced topics related to produce food safety and statistical modelling. Indirectly, the produce food industry will also benefit as they will be able to hire new personal that are better trained in produce food safety and statistical modelling. Changes/Problems:Goal 1. Nothing to report. Goal 2. We have experienced a delay in the activities planning for this Goal due to the universities' lockdown in response to the COVID-19 pandemic. However, we do not foresee this delay to extend through the next reporting period and no change will be required in our timeline or scope. Goal 3. Nothing to report. Goal 4. Nothing to report. Goal 5. Nothing to report. Goal 6. Nothing to report. What opportunities for training and professional development has the project provided?Goal 1. Two postdocs, Drs. Ece Bulut and Sarah Murphy have been trained in setup of a framework for sustainability assessment, framing of the risk questions and the development of farm-to-fork conceptual models. Additionally, an undergraduate student (Hannah Rosenthal) has been trained in produce food safety research, focusing on the consumer perceptions, beliefs and behaviors. Goal 2. This project goal has provided training opportunities in GIS and R based modeling tools for a post-doctoral researcher, PhD student and MS student who were supported on this grant. The post-doctoral researcher was offered a position at University of Florida, which is evidence of an increased skill set, and food safety training. Goal 3. Nothing to report. Goal 4. Nothing to report Goal 5. This project has already involved two PhD graduate students in the development of the produce safety certificate program course framework. The graduate students are integrally involved in the development of the 4 specialized produce safety tracks framework and content, as well as the advanced course requirements and content. The graduate students will continue to be integrally involved in the development of the course content and materials in the future years of the grant. Goal 6. One Research Associate Sr., Dr. Renato Orsi, and one technician, Zoe Wasserlauf, were trained in developing online surveys and were responsible for creating the first framework of the undergraduate-level produce food safety module. Two graduate students, Samantha Bolten and Alexandra Belias, were trained in developing the undergraduate-level produce food safety module slides and participated in the discussion-like recording of the first lecture in this module. How have the results been disseminated to communities of interest?Overall, results to date have been communicated with the project Advisory Council, which includes 20 members; so far this group has met 3 times. Details on specific communication that occurred for the different goals are provided below. Goal 1. Preliminary results have been shared among the members of the external advisory council (comprised of representatives of the produce industry - growers, processors and distributors, trade and research organizations, and government agencies). Additionally, preliminary results have been shared among researchers and students in the academic circles. Goal 2. Preliminary results have been communicated to the Advisory Council, and project collaborators. Future communications and presentations are planned for 2021, including submission of abstracts for the International Association of Food Protection, and draft manuscript for Virginia and Florida specific models. Goal 3. Nothing to report. Goal 4. Nothing to report. Goal 5. The extension and outreach certificate program plans have been shared with the Advisory Council. Goal 6. The first lecture of the undergraduate-level produce food safety module was shared among the other PIs and PDs as well as among all members of the Advisory Council. What do you plan to do during the next reporting period to accomplish the goals?Goal 1. Over the next reporting period, we will complete farm-to-fork conceptual models for E. coli O157:H7-lettuce and L. monocytogenes-cantaloupe commodity-pathogen pairs. We also plan to start building the quantitative microbial risk assessment (QMRA) models based on the elements of the completed conceptual models for each pathogen-food pair and start evaluating the other indicators of food safety sustainability (social, economic and environmental). Goal 2. Over the next reporting period, we will continue our analyses and draft manuscripts of the Virginia and Florida datasets, and other states/regions. We will also continue to develop models based on different levels of complexity to the produce industry; including, first level: temperature, (air/water), rainfall, other characteristics (e.g., water type); second level: first level plus physiochemical parameters (e.g., dissolved oxygen); third level: first and second level plus microbial test data; four level: first, second and third levels plus land-use data. These models range in the levels of difficulty for the stakeholders, for example, it is fairly easy to obtain weather data from public weather stations (first level), to increasing difficulty to monitor and obtain land-use data from GIS software. However, we hope to provide tutorials in the outreach objective. Goal 3. Four produce processing facilities with distinct layouts and geographic locations will be selected for the development of facility-specific Agent-Based Models (ABM). Facilities already participating on other projects involving the PDs and PIs may be selected. New facilities (i.e., facilities not involved in current parallel projects) will be visited by key member on this project to ensure they are appropriate for this goal. Longitudinal microbial data will be obtained from previous sampling efforts or new longitudinal samplings will be carried out if needed. ABM will start to be developed soon after the facilities are selected and the longitudinal data are obtained. Goal 4. Identification of existing models and modelling tools will be carried out in this next reporting period. We anticipate that a high proportion of the literature being used for Goal 1 will be valuable for Goal 4 as well. We also expect to create and apply a web-based survey to assess produce industry practices regarding the use of modeling tools and barriers to using them. Goal 5. During the next reporting period, the course content for the 4 specialized produce safety tracks will be generated. The basic introductory produce safety courses that all participants will be prioritized to be completed first. The subsequent specialized courses for the 4 tracks will be developed next, and finally the advanced courses. Goal 6. Over the next reporting period, we will finalize the undergraduate-level produce food safety module and develop the other three modules planned for this goal. We expect that both the undergraduate-level and graduate level produce food-safety modules will be finalized by the end of the next period while the statistical modelling modules should be initiated by then.

Impacts
What was accomplished under these goals? Goal 1. Progress was made on the assessment of sustainability of fresh produce food safety, specifically the assessment of the public health indicator of sustainability. Towards that goal, we are using the Modular Process Risk Model (MPRM) framework to model produce contamination prevalence and levels along the farm-to-table production chain. The models are being developed for two produce-pathogen pairs: (i) Escherichia coli O157:H7-lettuce (cut) and (ii) Listeria monocytogenes-cantaloupe (whole and cut), which were selected based on feedback from members of our external advisory council. A preliminary conceptual model was developed that outlines the risk pathway from farm-to-fork contamination chain for the E. coli O157:H7-lettuce commodity-pathogen pair. The E. coli O157:H7-lettuce model has been designed to estimate the number of E. coli O157:H7 cells per serving of chopped romaine lettuce and includes four consecutive modules: (i) growing, (ii) harvest/field packing, (iii) processing/distribution and (iv) consumer. A preliminary version of the E. coli O157:H7-lettuce conceptual model was presented to the project team and the external advisory council, and then updated according to their feedback. A sub-group was identified within the advisory council to further seek input on model design and parameterization (e.g., manure or treated biological soil amendment practices that are of specific interest to the produce industry and stakeholders). Accordingly, the main processes currently considered in the MPRM include (i) lettuce growth in the field, (ii) holding time, (iii) harvest/field packing, (iv) washing and chopping (post-harvest processing), (v) distribution, (vi) retail storage and display, (vii) transportation from retail to home, (viii) home storage, and (ix) consumer washing and preparation. Major considered sources of E.coli O157:H7 contamination in the lettuce MPRM are (i) irrigation water, (ii) soil and/or manure contamination, (iii) harvesting blades, (iv) hands of the workers and/or harvesting belts, (v) wash water (cross-contamination) and cutting equipment at post-harvest processing, (vi) and consumer washing (cross contamination) and preparation. Preliminary works has also started on the development of the conceptual model for the L. monocytogenes-cantaloupe MPRM. Goal 2. Due to COVID-19 restrictions, no sampling efforts were performed during this reporting period. To accomplish our target goal of using GIS-based modeling tools to reduce produce microbial food safety hazards in water, and environmental sources; we conducted a search of available datasets (both published and un-published) on the concentration of indicator organisms and presence of pathogens in surface water surfaces in the United States, and North America, to characterize and identify spatio-temporal factors (e.g., soil properties, meteorological events, and adjacent land-use), and build models to predict pathogen presence (e.g., Listeria monocytogenes, Salmonella), and generic Escherichia coli levels. Raw data was collected from 50 datasets and represented 22 states, 3 Canadian provinces, and an additional dataset from Mexico yielding 112,710 individual data points. All data was organized in Excel for importation into the statistical database R (to allow ease of storage, analysis and visualization). Additionally, R code was developed to mine adjacent land-use buffered at 122, 366, and 1098 meters from the sampled GPS location using the National Land Cover Database (NLCD). This was exacted for all data with GPS points (approximately 1/3). Historical weather data was also pulled from the nearest weather station using the date the sample data point was collected. For example, in Florida, 4 different datasets were used for mined, two of which are unpublished, and we performed analyses to determine associations between Salmonella presence, and indicator organisms (e.g., generic E. coli, fecal coliforms), and proximity to specific land-use categories (e.g., wetlands, cropland, forest, urban), meteorological variables (e.g., air temperature, rainfall), and physiochemical parameters (e.g., turbidity, conductivity). Similar analyses are underway for Virginia, and other states. Analyses are in progress. Goal 5. The development of a national produce safety certificate program that provides training and certification to individuals associated with all segments of the produce industry has been initiated. The courses that will make up the content for the 4 specialized certificate tracks, and the advanced certificate programs have been identified. The 4 specialized certificate tracks are Produce Growers, Produce Packing Houses & Produce Processors, Retail, and Consumer Educators. The basic produce safety courses that will be taken by all tracks include: Introduction to Food Microbiology, Systems Approach to Produce Safety, and Traceability in Produce Systems. For the Produce Growers specialized certificate, in addition to the basic produce safety courses, they will take courses that include: GAPs, and Control of Produce Contamination in the Field. For the Produce Packing Houses & Produce Processors, the courses identified for this specialized certificate will be: GMPs 21 CFR 117, PCQI or HACCP for Packing Houses, and Hygiene and Environmental monitoring. The courses for the specialized certificate for Retail will include: Food Code, and Additional Strategies for Retailers to Prevent Contamination. The Consumer Educators specialized certificate will include: Consumer Produce Safety for Educators. The advanced certificates for the 4 specialized tracks will require one advanced course after completing the course requirements for the respective specialized tracks. In addition, advanced courses for Modeling of Produce Safety and GIS Tools for Produce Safety will be developed as the respective objectives for modeling and GIS tools have been completed. Originally, the certificate program was going to be mixed model of on-line and in-person, but due to the pandemic, the certificate program will be on-line for the first 1-2 years, and may change to in-person and on-line when the pandemic subsides. As part of the industry training, we have also offered one industry webinar on WGS (live on November 24th; 50 registrants; 25 attendees), which specifically addressed the WGS data for the E. coli outbreaks and recalls that occurred during the fall of 2020. Goal 6. Progress was made on (i) assessing the needs for and (ii) initial development of the modules proposed under Goal 6. Briefly, a draft survey on needs assessment related to undergraduate and graduate-level modules to teach (i) produce food safety and (ii) statistical modelling was created and sent to three academic professors who teach food safety at three different US universities. Based on the feedback received from these three educators, a final survey was developed. The final survey had questions on format of the modules (e.g., number of lectures per module, length of each module) and as well as interest of using the modules. Educators responsible for teaching produce-related food safety lectures at 53 academic institutions across the United States were identified, contacted, and asked to complete the survey. Twenty out of the 57 educators contacted completed the survey after 22 days. We also obtained feedback from the Advisory Council during our second quarterly Advisory Council meeting, chiefly on the content of the modules. The surveys indicated that modules containing three 20-40 minutes lectures each would be ideal and most educators that responded the survey showed high interest in using the modules in their own classes, both as a slide set for in-person instruction as well as a recorded lecture for distance learning. We are currently developing of the undergraduate-level produce food-safety module. The first of the three lectures have been finalized and recorded using a discussion-like format (link).

Publications