THE RUMINANT FARM SYSTEMS (RUFAS) MODEL: A NEXT-GENERATION, WHOLE-FARM, DAIRY SYSTEMS MODEL TO SUPPORT SUSTAINABLE PRODUCTIVITY AND ENVIRONMENTAL HEALTH

• Sponsoring Institution: National Institute of Food and Agriculture
• Project Status: ACTIVE
• Funding Source: AFRI COMPETITIVE GRANT
• Reporting Frequency: Annual
• Accession No.: 1022804
• Grant No.: 2020-68014-31466
• Cumulative Award Amt.: $1,000,000.00
• Proposal No.: 2019-07851
• Project Start Date: Jul 1, 2020
• Project End Date: Jun 30, 2025
• Grant Year: 2020
• Program Code: [A1261]- Inter-Disciplinary Engagement in Animal Systems

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

Performing Department
Animal Science

Non Technical Summary
U.S. Dairy farms produce nutritious dairy products that help meet American requirements for protein, vitamins, and minerals but this production comes with significant environmental impacts. A simulation tool that allows scientists and dairy producers to compare the impact of management decisions and climate change on both production and environmental outcomes will support industry personnel and policy makers in their efforts to improve sustainability of the dairy industry. Thus, our objective is to build and test a computer simulation model - the Ruminant Farm Systems Model - that predicts milk and meat production, environmental impacts, and production costs of a dairy farm under common U.S. environment and management scenarios. Our final objective is to increase awareness and application of our model by members of the scientific and dairy communities by involving members of the dairy industry during the development and testing phases. We expect use of our model by scientists, producers, and policy makers to improve decision making surrounding sustainable dairy production resulting in continued supply of nutritious dairy products to the American people, improved economic viability for members of this industry, and reduced environmental impact from dairy farms.

Animal Health Component

50%

Research Effort Categories

Basic

50%

Applied

50%

Developmental

(N/A)

Classification

Knowledge Area (KA)	Subject of Investigation (SOI)	Field of Science (FOS)	Percent
307	3410	1060	60%
307	3410	1010	20%
307	3410	1070	20%

Knowledge Area
307 - Animal Management Systems;

Subject Of Investigation
3410 - Dairy cattle, live animal;

Field Of Science
1060 - Biology (whole systems); 1070 - Ecology; 1010 - Nutrition and metabolism;

Keywords

ruminant farm systems

precision animal management

environmental synergies of animal production

Goals / Objectives
Our long-term goal is to support sustainable profits, productivity, and environmental health in US dairy systems by making a modular, adaptable, open-source model available to both researchers and managers of real-world enterprises. We will create the Ruminant Farm Systems (RuFaS) model and support its dissemination through four objectives:1.Complete and integrate into RuFaS four biophysical process modules--Animal, Manure, Crop & Soil, and Feed Storage--that simulate production, nutrient cycling and loss, and resource use, including responses to changes in farm structure, management, and climate.2.Develop and integrate into RuFaS system balance modules that summarize impacts of management choices and biophysical module outputs on energy use and production, water use and pollution, and the relative costs of management choices.3. Calibrate and validate the RuFaS model using commercial farm data.4. Promote RuFaS model use and trust through demonstrations and training.

Project Methods
Throughout RuFaS development, we will emphasize modern coding practices that prioritize code clarity, documentation, future adaptability, and effective version-control strategies. Well-structured and well-documented code is essential for model transparency, intelligent, widespread application, and adaptation to future knowledge. The model will be open-source to encourage multiple, independent users and to allow them to make changes and updates to the code as necessary for their distinct purposes. To enable this, we have developed interdisciplinary and systematic model development methods as follows:1. Development of pseudocode and information flow charts. Subject-matter experts for each module will draw information from existing models and the literature to describe processes in a detailed, uniform format to serve as the foundation of RuFaS documentation. Description of and references for each simulated process will be included in the pseudocode, which, combined with multi-level information flow diagrams, will facilitate model clarity and use.2. Translation of pseudocode into code. Pseudocode and flow charts are translated into Python code by trained programmers in collaboration with discipline experts following agile development principles with two-week goals, objectives, and code reviews. Emphasis is placed on the clarity of code through code structure and logical and detailed commenting.3. Evaluation of sub-model, module, and model functions at increasing scales. As development progresses, we evaluate simulation processes ranging from single functions to modules to whole model performance using existing experimental and literature-based datasets. Datasets will stress the system where possible; in cases where adequate stresses do not exist, we will test model function at known biological boundaries.4. Evaluation of model structure and outputs for user application. As sub-modules and modules progress and are merged, simulation of model function will be demonstrated for stakeholders including the project's Industry Advisory Board (IAB), extension networks, and research community. Solicited feedback is incorporated into model development.Existing progress provides confidence that our model development strategy and data management plan will produce clear documentation and a readable, readily adaptable codebase. This will encourage model implementation and adaptation by future users with diverse backgrounds, which is essential to model evolution and longevity.Data Analysis.During the first round of validation, we will assess prediction accuracy and conduct a sensitivity and uncertainty analysis. We will use the results of this initial validation to identify parts of the model that need improvement and with high uncertainty. To assess prediction accuracy, we will compare model predictions with values commonly collected on farm like milk production, crop production, manure production, and feed imports to create empirical estimates of nutrient imports, exports, and internal flows. Model evaluation and sensitivity analysis techniques will be implemented in Python and R.Evaluation.Project objectives will be evaluated for success in three areas: (1) model function, (2) model suitability, and (3) model distribution and application.Model Function. The primary indicator of success will be a fully functional and tested Version 1.0 of the model. By the end of Year 2, V1.0 of RuFaS will be ready for use and evaluation by the project members. To develop the model codebase, agile model development processes will be coordinated and implemented across teams. The agile work plan includes bi-weekly meetings with active programmers, quarterly conference calls with team leaders, and annual investigator and IAB team meetings to assess progress. Team conference calls are already occurring, and we held three annual team meetings in 2017, 2018, and 2019.Model Suitability. It is essential that RuFaS be able to simulate commonly used and emerging management strategies and technologies in a way that is directly relevant to the dairy industry. For this reason we will engage an Industry Advisory Board during the model development and testing phase through participatory modeling efforts.Model Distribution and Application. RuFaS progress will be widely advertised and demonstrated at technical conferences, extension meetings, workshops, and on-farm visits both during development and upon completion. Specific extension-focused opportunities to promote RuFaS during and after the project include the annual Cornell Nutrition Conference, Four-State (WI, MN, IA, IL) Nutrition and Management Conference, Cornell's Dairy Nutrition Short courses, and UW-Madison's Midwest Manure Summit, among other conferences hosted by Cornell's Pro-Dairy and UW- Madison's extension networks.Criteria for RuFaS outcomes evaluation will also include (1) meeting attendance and minutes to assess extent of collaboration across government, industry, and academic institutions and participation by women and underrepresented groups, (2) evaluation of GRA and post-doc progress and success, and (3) size, type, and feedback of audiences exposed to RuFaS, measured through presentations, publications, extension activities, and coursework.

Progress 07/01/22 to 06/30/23

Outputs
Target Audience:We continued to engage our Stakeholder Advisory Council which is composed of dairy producers, non-profit organizations, and dairy industry professionals. In addition we presented our work at several dairy industry and academic conferences to share our work with a wider audience of dairy industry professionals and the scientific agricultural modeling community. Changes/Problems: We again had trouble finding the right personnel to meet our technical needs. The post-doc for the Crop and Soil module only stayed with us for 4 months and so are again searching for someone to fill that role. Although we were able to find a graduate student to work part time on the systems balance modules, we are still lacking in support for this part of the model. What opportunities for training and professional development has the project provided? Postdoctoral and graduate students attended both the RuFaS Annual meeting and at least one academic conference We hosted an informal educational meeting series titled 'Introduction to Machine Learning' that was open to all project members How have the results been disseminated to communities of interest? Nothing Reported What do you plan to do during the next reporting period to accomplish the goals? The FARM-ES model that is a whole farm footprinting tool will launch its version 3 which will use the RuFaS model as its backend. We will make our Github repository open to the public We will complete the evaluation of all parts of the biophysical modules We will complete the systems balance modules

Impacts
What was accomplished under these goals? Complete and integrate into RuFaS fourbiophysical process modules--Animal, Manure, Crop & Soil, and Feed Storage--that simulate production, nutrient cycling and loss, and resource use, including responses to changes in farm structure, management, and climate. We made significant progress on the models' methods for input data collection and output data generation during this period. By improving these methods, we made the model much easier to work with and for subject matter experts to engage and review. With a change in personnel, we decided to completely revise the soil and crop module to improve the modularity of the codebase and documentation. With the improved coding methods and additional software engineer support, we were able to replace all existing functionality and expand the field management methods in just over 8 months. Develop and integrate into RuFaSsystem balance modulesthat summarize impacts of management choices and biophysical module outputs on energy use and production, water use and pollution, and the relative costs of management choices. A graduate student was brought on to develop the systems balance methods and initiated the development of the methodology to estimate the energy required for field and manure management. We collected a database of feed prices with which we can initialize the model to support economic inference into feed use. Calibrate and validatethe RuFaS model using commercial farm data We collected RuFaS input values from 32 pilot testing farms and initiated simulation of these farms PromoteRuFaS model use and trust through demonstrations and training. We presented RuFaS progress updates at 3 industry conferences and 4 academic conferences?

Publications

• Type: Journal Articles Status: Accepted Year Published: 2023 Citation: Li, M., K.F. Reed, M.R. Lauber, P.M. Fricke, and V.E. Cabrera. 2023. A stochastic animal life cycle simulation model for a whole dairy farm system model: Assessing the value of combined heifer and lactating dairy cow reproductive management programs. J. Dairy Sci. 106: J. Dairy Sci. 106:3246-3267. https://doi.org/10.3168/jds.2022-22396
• Type: Journal Articles Status: Published Year Published: 2023 Citation: Li, M., K.F. Reed, V.E. Cabrera. 2023. A time series analysis of milk productivity in US dairy states. J. Dairy Sci. 106:6232-6248. https://doi.org/10.3168/jds.2022-22751.
• Type: Websites Status: Other Year Published: 2023 Citation: rufas.org

Progress 07/01/21 to 06/30/22

Outputs
Target Audience:Our target audiences are: Animal and Agricultural Systems Scientists Organizations developing farm-scale footprinting tools Dairy industry professionals and non-profit organizations working to reduce the environmental footprint of dairy production Federal and State governments defining regulations related to GHG emissions and nutrient losses from dairy production Changes/Problems:We continue to struggle to find a suitable graduate student to work on the systems balance modules and after less than a year, lost the post-doctoral scholar who was to continue work on the soil and crop module so will need to replace them before work can continue in earnest on that module again. What opportunities for training and professional development has the project provided?- Participation of student software engineers in the model development provides these students with opportunities for them to apply their skills sets in a real world setting. - The post-doctoral scholars were given opportunities to participate in Stakeholder Advisory Council meetings to learn to communicate with Stakeholders and implications of applied research. How have the results been disseminated to communities of interest?We presented updates at 5 Industry facing conferences as well as 2 abstracts at the American Dairy Science Association's Annual meeting. What do you plan to do during the next reporting period to accomplish the goals?We plan to continue developing the functionality of the 4 biophysical modules and make progress on model evaluation.

Impacts
What was accomplished under these goals? 1. We made significant progress on the Manure module during this period and further expanded functionality of the Animal Module including refinement of the integration between the animal life cycle and nutrition functionalities. 3. We initiated evaluation of the animal module including assessment of the impacts of reproductive management on economic and environmental outcomes. Preliminary evaluation of the Crop and Soil module was initiated with datasets collected at USDA-ARS and university sites in Wisconsin. Additional datasets were gathered for evaluation of this module from USDA LTAR sites. 4. We presented updates at 5 Industry facing conferences as well as 2 abstracts at the American Dairy Science Association's Annual meeting.

Publications

• Type: Journal Articles Status: Published Year Published: 2022 Citation: Li, M., G.J.M. Rosa, K.F. Reed, V.E. Cabrera. 2022. Investigating the effect of temporal, geographic, and management factors on US Holstein lactation curve parameters. J. Dairy Sci. 105:7525-7538. https://doi.org/10.3168/jds.2022-21882

Progress 07/01/20 to 06/30/21

Outputs
Target Audience:The target audience includes scientists and professionals in the agricultural sciences and the dairy industry. Changes/Problems:The primary challenge that we have faced has been finding qualified candidates to meet our needs for biological systems modelers. In particular, we have not been able to find suitable candidates for the Post-doctoral scholar to lead development of the Crop and Soil module or a PhD student to work on development and testing of the ration formulation part of the animal module. We are continuing to seek candidates for the Crop and Soil module but have chosen to use some of the funds allocated to a PhD student in Animal Science to hire a professional programmer who can act as a mentor to the undergraduate student software engineers. What opportunities for training and professional development has the project provided?This project has provided an opportunity for 9 undergraduate students from computer science and engineering backgrounds to work with an interdisciplinary team on a project with real world applications and impact. These student programmers have opportunities to learn new programming languages and skills and can move into leadership roles within sub-teams. In addition, the Cornell team visited one commercial dairy farm in this reporting period to give the students from non-agricultural backgrounds an opportunity to see and interact with farmers and their livestock. Three PhD students and 4 post-docs have also contributed to our work. They have benefited from regular meetings with the project team mentors, participation in executive and industry meetings, presentation of progress updates at project meetings and scientific conferences, and development of leadership skills gained through sub-team management. How have the results been disseminated to communities of interest?As we reported on our progress towards our Objective 4 above, we have: hosted 4 meetings with our Industry Advisory Council hosted a 3-day workshop in March of 2021 in which scientists outside of the direct RuFaS project were invited to learn about and use RuFaS for their research purposes. This workshop included participation of 2 USDA-ARS scientists, 2 US faculty, 2 postdocs, 1 graduate student and 1 international scientist from outside of the RuFaS team. Presented 3 abstracts at the American Dairy Science Association Annual Meeting Introduced the project and its justification at the Cornell PRO-Dairy Nutrition Short Course in June 2021 In addition, we published two manuscripts listed in our products and established the RuFaS website: rufas.org. What do you plan to do during the next reporting period to accomplish the goals?In the next reporting period we plan to complete the first version of the fully-connected dairy farm model by: Completing translation of the manure module pseudocode into the RuFaS python code and merging with the rest of the model. Updating the Feed storage model to a dynamic, daily time-step model based on the existing pseudocode Continuing to test individual processes and sub-modules for directional accuracy and precision with experimental data We will build the system balance modules within the RuFaS code base by translating the literature data on feed prices and management costs into SQLite databases and creating Python modules to summarize management costs and environmental impacts. We will also begin the process of testing the RuFaS model on commercial farms by initializing a pilot-testing program to establish baseline footprint estimates and compare the impact of a subset of potential management scenarios. In addition to the above, we will continue to hold our regular agile development meetings, executive and industry advisory council meetings and organize an annual meeting workshop to share progress and set the trajectory for the coming year.

Impacts
What was accomplished under these goals? The Ruminant Farm Systems model will fill a gap in the need for farm system level decision support in dairy production. By building a flexible, well-documented and open-source model that simulates nutrient cycling, production, and emissions on a dairy farm, RuFaS will increase our understanding of downstream impacts of management decisions, enable whole farm assessment of environmental impacts, and evaluation of management scenarios for their expected impact on production, GHG emissions, water use and quality, and management costs. Application of RuFaS will benefit (1) scientists in their ability to assess the whole farm implications of new technologies and management practices, (2) dairy producers and industry consultants in through better informed management decisions, and (3) industry organizations and policy makers in their development of regulatory and incentive programs. Our long-term goal is to support sustainable profits, productivity, and environmental health in US dairy systems by making a modular, adaptable, open-source model available to both researchers and managers of real-world enterprises. Objective 1: Complete and integrate into RuFaS four biophysical process modules--Animal, Manure, Crop & Soil, and Feed Storage--that simulate production, nutrient cycling and loss, and resource use, including responses to changes in farm structure, management, and climate. We have made significant progress towards our goal of completing a fully integrated dairy farm systems model through our agile development methods. In particular, we added the following pseudocode to our documentation: Soil Carbon cycling Crop growth for grass, winter wheat, rye and beets Nutrient requirements for calves, growing animals, and dry cows Non-linear least-cost ration optimization Maintenance of P and N balance for individual animals Manure management system structure plus nutrient transformation and emissions under: Flushing and scraping manure collection Long-term covered and uncovered lagoon storage Anaerobic digestion Dynamic nutrient transformation and loss during forage drying and ensiling Of the above, we have completely integrated #1-4 into the model and making progress on #5-7. We have added thousands of lines of code, improved the program modularity and modified the input-output structure to improve the ease of use. Objective 2: Develop and integrate into RuFaS system balance modules that summarize impacts of management choices and biophysical module outputs on energy use and production, water use and pollution, and the relative costs of management choices. We have held regular Systems Balance sub-team meetings that have resulted in significant progress towards developing the methods to summarize and estimate the whole farm nutrient cycling, GHG emissions, energy use and production, water use, and management costs. In particular, we have collected all the necessary supporting equations from the literature in order to estimate farm electricity and fossil fuel use as well as the embedded C footprint in purchased fertilizer. In addition, we have made significant progress in developing a feed price database and price databases for the cost of electricity and water. Finally, we have initiated work on development of a database for the embedded carbon and water footprint in purchased feeds. Objective 3: Calibrate and validate the RuFaS model using commercial farm data. We have not started the process of evaluating the model through simulation of commercial farms but we have taken steps to prepare for this process. In particular, we built a user-friendly data collection web application that automates data importing into the RuFaS JSON input file format and we have initiated discussions with industry members and our Industry Advisory Council to select commercial farms for participation in our program. Objective 4: Promote RuFaS model use and trust through demonstrations and training. We have hosted 4 meetings with our Industry Advisory Council which included demonstration of new model functionality and participatory modeling discussions to gain input into development direction. Thus, we are promoting target audience trust in the RuFaS model through active stakeholder engagement in the model development and testing process. In addition, we hosted a 3-day workshop in March of 2021 in which scientists outside of the direct RuFaS project were invited to learn about and use RuFaS for their research purposes. This workshop included participation of 2 USDA-ARS scientists, 2 US faculty, 2 postdocs, 1 graduate student and 1 international scientist from outside of the RuFaS team. In addition to meetings and workshops hosted by the RuFaS team, we have also shared our work with the scientific community at the American Dairy Science Association Annual Meeting (3 abstracts presented in 2020) and the Cornell PRO-Dairy Nutrition Short Course in June 2021. During this reporting period, we have made progress towards development of our key output which is the whole farm simulation model. In addition, we have built a foundation of trust and engagement with our industry collaborators that has increased awareness and understanding of the essential role that models can play in finding systems solutions to challenges facing the dairy industry.

Publications

• Type: Journal Articles Status: Published Year Published: 2021 Citation: Hansen, T.L., M. Li, J. Li, C.J. Vankerhove, M.A. Sortirova, J.M. Tricarico, V.E. Cabrera, E. Kebreab, K.F. Reed. 2021. The Ruminant Farm Systems Animal Module: A biophysical description of animal management. Animals. 11:1373. https://doi.org/10.3390/ani11051373.
• Type: Journal Articles Status: Under Review Year Published: 2021 Citation: Li, J.H., E. Kebreab, F. You, J.G. Fadel, T.L. Hansen, C. VanKerhove, and K.F. Reed. (minor revisions returned 9/7/2021). The application of nonlinear programming on designing feed formulation for dairy cattle. J Dairy Sci.