Progress 01/30/23 to 01/29/24
Outputs
Target Audience:We developed a list of 75 stakeholders, all of whom were invited to the first engagement event, held on December 5, 2023, in Charlotte, alongside the Sustainable Ag Summit (Dec 6-7), which many of these stakeholders generally attend. The in-person engagement event was entitled "Managing Shocks in the Animal Protein System" and was attended by around 25 stakeholders. The stakeholder list follows: Ellen Lai (ABS); Janice Rueda (ADM); Scott Higgins (American Dairy Association Mideast); Sarah Novak (American Feed Industry Association); Keith Dillon (ASN); Gwen Twillman (ASN); R Cook (Bamert Seed); Jeff Clark (C-Lock Inc.); Michelle Grogg (Cargill); Amanda Bushell (Context Network); Sarah Frank (Context Network); Matt Sutton-Vermeulen (Context Network); Kelsey King (Dairy Farmers of America); Michelle Schack (Dairy Kind); Rashel Clark (Dairy West); Andrew Coffey (DMI); Lisa McComb (DMI); Alyssa Sundell (DMI); Karen Scanlon (DMI & Innovation Center for US Dairy); LaKisha Odom (FFAR); Meredith Ellis (G Bar C Ranch); Donald Moore (GDF); Angela Ailloni (Ginkgo Bioworks); Ashley McDonald (Grasslands Coalition); Abby Bauer (Hoard's Dairyman/J of Nutr Mgt); Rick Naerebout (Idaho Dairyman's Association); Lara Moody (IFEEDR); Sylvia Rowe (Independent Scientist); Angela Hessinger (Innovation Center for US Dairy); Sarah Klopatek (JBS); Sarah Alexander (Keystone); Jonathan Geurts (Keystone); Brad Sperber (Keystone); Cynthia French (LR Bioenergy); Jenni Tilton-Flood (Maine Dairy Promotion Board/Maine Dairy Nutrition Council); Jordan Sabine (McDonald's); Julie Mattson Ostrow (Midwest Dairy); Lauren Servick (Minnesota Pork); Mitch Breunig (Mystic Valley Dairy); Jamie Burr (National Pork Board); Marguerite Tan (National Pork Board); Jeff Piggot (National Pork Producers Council); Dan Raiten (NIH); Shelly Mayer (PDPW); Nicki Briggs (Perfect Day, Inc.); Bill Salas (Regrow); Klaus Kraemer (Sight & Life); Chris Gambino (The Breakthrough Institute); Sasha Gennet (TNC); Clare Kazanski (TNC); Naglaa El-abbadi (Tufts U); Justin Ransom (Tyson Foods); Joanne Slavin (U Minnesota); Pedro Urriola (U Minnesota); Natalie Lounsberry (U New Hampshire); Frank Mitloehner (UC Davis); Jolene Griffin (United Dairy Industry of Michigan); Janice Giddens (US Dairy Export Council); Kelly Sheridan (US Dairy Export Council); Ying Wang (US Farmers & Ranchers in Action); Rebecca Bolton (US Roundtable for Sustainable Beef); Megan Farley (US Roundtable for Sustainable Beef); Samantha Werth (US Roundtable for Sustainable Beef); Diane DeBernardo (USDA); Naomi Fukagawa (USDA); Mike Grusak (USDA); Gregory Bohrer (Walmart Foundation); Emily Moberg (WWF); Shari Westerfield (Zoetis). Changes/Problems:UMN has had some project staffing challenges and consequential changes. UMN had originally had a life cycle analysis scientist (Dr. Rylie Pelton) included on this project. However, after the project was funded, and before it started, Dr. Pelton's employment effort at UMN went from 100% to 25%, drastically limiting her capacity to participate in this project. She is now only playing a minor advisory role. Furthermore, our original budget included funds for a software engineer who we had employed at the time of grant submission, but who had left before this grant began. Given the status of our team and the FoodS3 model at the time the grant started we pivoted from a software engineer to bringing on a postdoc. We began the process for hiring in March 2023, selected Dr. Dhar in August 2023 and he was only able to start at the end of November 2023. What opportunities for training and professional development has the project provided?UMN was able to hire a postdoctoral researcher (Dr. Aurup Ratan Dhar) in late November 2023 to join this project. While Dr. Dhar's time during the first year of the grant was minimal, we are excited to have his expertise on nitrogen and phosphorus footprinting in agri-food systems. Dr. Dhar's previous focus has been the Indian Subcontinent, and we are working to familiarize Dr. Dhar with the relevant agricultural and environmental impact data in the US. Stanford provided an opportunity to three Master's students to conduct a quarter-long research project on the prediction of animal farm headcount from satellite imagery. Dr. Lobell met with them weekly, and the work was written up and submitted to the ICML conference. How have the results been disseminated to communities of interest?The team held a December 5 stakeholder engagement event, entitled "Managing Shocks in the Animal Protein System," alongside the Sustainable Ag Summit (Dec 6-7) in Charlotte. What do you plan to do during the next reporting period to accomplish the goals?Objective 1: As described in the project timetable, the team will host a second stakeholder engagement event, most likely again alongside the Sustainable Ag Summit, this year very conveniently located in Minneapolis (November 20-21). The primary purpose of this event will be to collect stakeholder feedback on the types of interventions that would be of greatest interest for us to include in the project. Objective 2: We will finalize our facility optimization (Task 2.1) and continue working on the monthly timestep data (Task 2.2), that will include bringing in the USDA's 2022 Ag Census data that was released in Feb 2024. We will also begin gathering the data required for estimating county milk consumption (Task 2.3). The prototype system dynamics model was well received by stakeholders, who encouraged the team to continue developing it. It was decided that, in addition to the value the simpler modeling framework brings to stakeholder engagement, the prototype approach would also be a powerful tool for guiding the implementation of dynamic elements in FoodS3. Therefore, further development of the prototype will continue in parallel with the work on FoodS3. Enhancements will include expansion of the model from its current national scale to a multi-regional scale, as well as addition of environmental impacts linked to the core supply chain. Objective 3: We will hire two summer undergraduate students to build out training data for being able to use satellite data and ML to detect hog and dairy manure management systems. With this dataset, we will then begin work on using satellite data to characterize the manure management systems for dairy and hog farms across the US (Task 3.1). We will also continue to monitor developments in this space by other groups, such as the new dataset of emissions sources released by ClimateTrace, to see if these are suitable inputs to the Foods3 model. We will also decide if we will be able to feasibly use remote sensing to estimate on field lime application amounts in a manner that can help us spatialize the greenhouse gas impacts at the county level (Task 3.2). Objective 4: We will begin looking into relevant data sources on food affordability (Task 4.1), jobs and income (Task 4.2), and nutrient availability (Task 4.3). Objective 5: This work will begin in 2025.
Impacts
What was accomplished under these goals?
The team held a kickoff workshop on March 21-23 at the offices of the University of Minnesota Institute on the Environment to develop a plan for the first year of the project. Biweekly one-hour zooms were used by the team to monitor progress. Objective 1: It was decided that the best way to engage stakeholders would be to develop a prototype dynamic model of the beef system, to help stakeholders visualize what would eventually be possible with an enhanced version of the FoodS3 model, as well as gathering feedback on what types of shocks and interventions would be of most interest. The prototype model, developed using the Stella system dynamics software package, consists of stocks and flows graphically representing the key stages of the US beef supply chain. The model dynamically tracks the movement of animals and animal products through the supply chainand includes the following activities: Breeding and pregnancy Raising calves stocker and finishing of cattle slaughterhouse processing consumer supply and demand dynamics The model was demonstrated in real time during the stakeholder meeting. Stakeholders had an opportunity to observe the dynamic response of the beef system to disease shocks at various stages of the supply chain. They could observe how decision makers at each point in the model supply chain respond to price supply and demand dynamics that flow through the supply chain from the consumer to the rancher and from the rancher back to the consumer. Running the model in real time allowed stakeholders to see the boom-and-bust cycles that can continue for years after the original disruption has ended. They immediately recognized and confirmed the kinds of patterns the model predicted. Seeing the model run in real time and walking through the graphical representation and logic of the US beef system model sparked a lively discussion with stakeholders, who were able to generate ideas for improvements as well as priorities for what kinds of disruptions they felt were most important to evaluate. We developed a list of 75 stakeholders, all of whom were invited to the first engagement event, held on December 5 in Charlotte, alongside the Sustainable Ag Summit (Dec 6-7), which many of these stakeholders generally attend. The in-person engagement event was entitled "Managing Shocks in the Animal Protein System" and was attended by around 25 stakeholders. Very helpful feedback was received at the event and is briefly summarized below: Stakeholder feedback on shocks: Seed availability shocks (corn, etc.) Stockers Demand shocks (e.g., boycotts, fear of BSE, etc.) Import/Export bans Fires at packing houses Labor/immigration Forensic analysis of past 30 years to find others Key inputs, e.g., vitamins from China Port strikes Policy changes, e.g., pricing methane, C markets, climate-smart commodities, etc. Include "Negative shocks" - e.g., disruptive tech that boosts productivity or eliminates bottlenecks Stakeholder feedback on modeling: Ensure model has decision points related to actions that can build system resilience Add stocker backgrounders Add decision factors other than just price Add feedlot operators, stockers, auction barns, packing facilities etc. Vertically-integrated supply chains Interaction between shocks? Stakeholder feedback on methodology Regionalization: Use NCBA regions, or possibly states; More localized pricing needed? Timestep: Quarterly seems fine for beef; Monthly may be needed for pork & poultry Metrics: Make sure to include metrics of interest to decision-makers, e.g., protein per capita Objective 2: Task 2.1: We have been working to update the beef, broiler, hog, and raw milk primary processing facility capacities to an annual timestep. Our previous work had facility capacities estimated every five years using a combination of datasets from industry reports detailing total production output by company and/or capacity depending on the livestock category, USDA ERS reports detailing the state and national total slaughter by livestock category, and satellite imaging to identify relative sizes of facilities in order to allocate between facilities. To standardize our methodology across animal facility types, we have improved our facility capacity estimation by aligning our data types for company data, standardizing the usage of USDA and Reference USA data, and using a consistent linear optimization to fill gaps in data. We are finalizing this update and working on a data paper outlining this new methodology. We have also begun work on building an annual time series for facility capacity using this new method. We also explored the idea of using imagery from Planet satellites to monitor individual locations. However, Planet satellites only see each facility twice a day. At best this data could provide a sense of monthly variation in slaughter capacities. We have decided to revisit this possibility later in the study after updating using non-spatial data first. Task 2.2: We are also working to "reconstruct the FoodS3 commodity flow transport optimization model's foundational structure to use monthly input and output information at livestock production, processing, and consumption nodes, to provide monthly origin-destination estimates at the commodity level." Toward this task we have assessed which data for which commodities is available at the monthly time-step in USDA NASS. We are currently assessing similar data in other USDA datasets (ERS, ARMS). Objective 3: Task 3:1: A team of students and a research assistant at Stanford University worked under the supervision of Dr. Lobell to develop models to estimate the number of animals in dairy, beef, and poultry farms. The model was primarily developed with data in California but subsequently tested in other states, showing reasonable performance. Our initial work focused on estimating headcount as some ground-based estimates were available for training and validation. Given the success of the aforementioned work we are optimistic that given appropriate training data, we can use ML to identify various manure management types at hog and dairy farms across the country, as well as to estimate the number of animals being delivered to slaughter facilities. We will be hiring students this summer to begin building the training data. Task 3.2: In FoodS3, over 80% of the full cradle-to-farm gate life cycle GHG emissions from crop production have been estimated using spatially explicit county-scale data for corn, soy, winter wheat, spring wheat, and durum wheat. A key remaining emission source that is likely to vary substantially across production regions is Lime application. However publicly available lime data is only available at the national scale. Our team has reached out to two major agribusinesses to see if they are willing to share their lime application data. Our team has looked at planet imagery in Georgia to see if the brightened visual signature of lime application is visible. We believe we have found examples of it, so it is possible an unsupervised ML model could work, but ideally have some ground truth data from an agribusiness. Objective 4: Nothing to report Objective 5: Nothing to report
Publications
|