Source: PURDUE UNIVERSITY submitted to NRP
AGENT BASED MODELING TO IMPROVE CAGE-FREE HOUSING SYSTEMS: WHAT DOES THE HEN SEE?
Sponsoring Institution
National Institute of Food and Agriculture
Project Status
ACTIVE
Funding Source
Reporting Frequency
Annual
Accession No.
1028296
Grant No.
2022-68014-36666
Cumulative Award Amt.
$1,000,000.00
Proposal No.
2021-10915
Multistate No.
(N/A)
Project Start Date
Feb 1, 2022
Project End Date
Jan 31, 2027
Grant Year
2022
Program Code
[A1261]- Inter-Disciplinary Engagement in Animal Systems
Recipient Organization
PURDUE UNIVERSITY
(N/A)
WEST LAFAYETTE,IN 47907
Performing Department
Animal Science
Non Technical Summary
Egg production in the US is changing from caged to cage-free production, with retailers and food service companies making a commitment to use only cage-free eggs. This requires that 72% of laying hens are raised in cage-free housing by 2025. However, there are many different cage-free designs as and how hens perceive and perform in these different designs is unknown. As more hens are being housed in cage-free housing systems, it is important to design systems based on hens' perceptions and needs, rather than on human perceptions. This project aims to understand how hens see the environment so that new cage-free housing can be created that positively impacts hen behavior, productivity and welfare, and food safety. This project will 1) collect information on the laying hens' eyes, physiology, behavior and welfare in two cage-free housing systems; 2) develop models to estimate how hens see their environment; 3) develop Agent Based Models (ABM) for testing virtual cage-free housing systems; and 4) build and validate the virtual cage-free system in real life. This approach will facilitate the design of a housing system better suited to the laying hen by reducing unwanted behaviors and improving hen welfare. Egg producers will be able to understand hens' perceptions and use of the cage-free environment, guiding producers' decision-making regarding hen housing and management. Additionally, housing manufacturers will have a new approach that could be used to design future cage-free housing systems.
Animal Health Component
40%
Research Effort Categories
Basic
10%
Applied
40%
Developmental
50%
Classification

Knowledge Area (KA)Subject of Investigation (SOI)Field of Science (FOS)Percent
30732101060100%
Goals / Objectives
The working hypothesis is that by quantitatively considering how hens see their world, we will have the ability to maximize/minimize the perception of different cage-free components (e.g., drinkers, perches, conspecifics, etc.) and thus explore barn spatial configurations that lead to positive changes in hen behavior to optimize productivity and welfare.Specific AimsCollect baseline information on laying hens' visual systems, physiology, behavior, welfare, and production outcomes in two types of existing commercial cage-free housing systems.Develop visual models to quantitatively estimate how hens perceive housing environments under different light conditions.3. Develop agent based models of virtual designs for new cage-free housing systems.4. Validate the next-generation virtual cage-free housing system in an empirical setting.
Project Methods
SA1: The visual models require specific types of information: visual parameters of the hens, reflectance properties of objects as well as the visual backgrounds of those objects, and spectral properties and intensity of the lighting systems. Reflectance of environmental objects and the visual backgrounds and irradiance of light will be measured. Additional measures of the spectral properties of the light in different locations within the housing system will also be collected in commonly utilized commercial lighting conditions. The visual traits of hens living in the same facilities will also be characterized from white feathered laying hens transported to Purdue University for sample collection.Experimental Design 1. White feathered laying hens will be raised in two different types of cage-free housing systems: floor and cage-free aviary. A single room of each housing system at the Purdue laying hen research facility will be used. The lay cycle (17 to 85 weeks of age) will be a typical single-cycle flock. Hens will be sampled throughout the production cycle and replacement hens will be added to maintain the stocking density. Room performance data will be collected including egg production, egg weights, feed consumption, water consumption, body weight, and livability throughout the production cycle of the laying hens. The welfare quality assessment of hens in the cage-free systems will be evaluated using an internationally accepted standard (Welfare Quality® (WQ)). WQ will be used to assess hens throughout the production cycle. Hens will be sampled at four different ages from the each of the two housing systems, coinciding with behavior sampling and tracking data collection. Information collected from the hens will include retinal physiology as pertains to visual perception and brain photoreception as pertains to gonadal function. Eye axial, corneal, and transverse measurements will be taken. Retinal photographs in both brightfield and epiflourescent light will be collected. Cone densities will be estimated by identifying the retinal oil droplets associated with each photoreceptor type. Based on the shape of the absorbance curves during collection, different types of cone photoreceptors associated with the visual pigment and oil droplet types will be distinguished. Neuroendocrine markers for reproductive activity will be assessed using qRT-PCR. Brains will be microdissected for discrete populations of neuroendocrine markers of welfare. Hens will have blood and tissues for endocrine indicators of stress and welfare collected. Finally, the same hens will be assessed for ovarian follicle numbers. Videos will be recorded to track bird location and examine bird behavior. Behavioral data will be extracted from video. A custom map of each housing system to enable hen location with x, y coordinates to be determined (for the cage-free aviary system that has multiple tiers, each tier (z) will have unique x, y coordinates). Hen location will then be used to create movement rules for the ABM. Quality of nest box and mislaid eggs will be assessed over the course of the flock cycle. The evaluation will include: volume of shell, static compression shell strength, shell thickness, Haugh unit, yolk index, vitelline membrane strength and elasticity, and whole egg total solids. Fresh fecal pools will be collected and sent forSalmonellaspp. isolation. Egg shell and egg contents pools will be collected forSalmonellaspp. detection. All samples will be assessed for prevalence ofSalmonellaspp. according to common laboratory practices.SA2: Visual contrast models take into account four parameters: (1) the reflectance spectra of the object of interest, (2) the reflectance spectra of the background, (3) the properties of the ambient light environment and (4) different visual traits of the target species. All of these parameters will be measured in SA 1a in commercial and research cage-free systems. The chromatic and achromatic contrast of the housing elements (perches, ramps, drinkers, etc.) under the light conditions of the housing facilities (different LED, etc.) will be modeled. Visual acuity models will be used to assess how hens resolve the visual spatial patterns of objects at different distances and under different light conditions. Visual acuity is a function of eye size and the density of the retinal cell mosaic (e.g., cone photoreceptors84), which will be measured in SA 1. Pictures of the inanimate and animate objects modeled in the visual contrast section above will be processed via AcuityView. Use the visual model developed for hens and explore, via simulations, the optimal lighting conditions to enhance/reduce the perception of different objects in the housing system.SA3: An ABM that simulates the lives of laying hens in cage-free systems from 20 to 85 weeks of age (single production cycle) will be built. The ABM will be parameterized with the baseline behavioral rules that hens followed when interacting with each other and the housing environment and with values related to the visual perception of hens. The ABM will be calibrated by iteratively revising parameter estimates until patterns emerging from the model are independent of rules imposed by model structure and conform to empirical observations. The relative animal welfare and rate of mislaid eggs will be compared across the simulated scenarios to determine expected performance of novel designs for cage-free housing. The range of factors and combinations of factors to be emulated for generation and testing will include scenarios that alter the configuration and arrangement of elements (e.g. nest boxes, feeders, etc.) of the housing system, the levels and types of light present in the housing system, the color of elements of the housing system (e.g. perches, drinkers, etc.), the overall dimensions of the housing system, and the density of birds within the housing system. Predictions and sensitivity analyses from the highest performing scenarios will be summarized and a survey developed to be administered to animal welfare NGOs and the Advisory Board to prioritize the various outcomes.SA4: White feathered laying hens will be raised in three different types of cage-free housing systems. The floor and cage-free aviary systems will be replicated from SA 1 and compared to the "best" designed system identified in SA 3. The "best" designed cage-free system will be constructed at the Purdue laying hen research facility where the other two systems are located. Hen will be housed at industry cage-free density and maintained through a lay cycle (17 to 85 weeks of age). Performance data will be collected throughout the lay cycle including egg production, egg weights, feed consumption, water consumption, body weight, and livability in the housing systems. Physiological, behavioral welfare data, egg quality and Salmonella shedding will be collected in all three housing systems as outlined in SA1. The Pattern Oriented Modeling (POM) procedure will be utilized to revise the design of the ABM to improve the fit of its predictions to observed patterns following the assessments of the performance and welfare of theredesignedcage-free system.

Progress 02/01/24 to 01/31/25

Outputs
Target Audience:Researchers, academic audiences and stakeholder advisory group Changes/Problems:The ABM is taking longer than anticipated to complete, however, Dr. Cain is confident everything will get done in time. The commerical industry survey has been delayed due to HPAI however, it will be disseminated after the Easter holiday. What opportunities for training and professional development has the project provided?Dr. Cain participated in a workshop on learning how to use the CTMM R package to analyze movement patterns.The project has provided opportunities to train a graduate student in experimental design, data collection, data analysis, scientific writing, and supervision and mentoring of undergraduate students. The graduate student has also received cross-disciplinary training in poultry science, animal welfare science and computer science. In addition, undergraduate students were trained in animal behavior and welfare data collection and using different types of technology to analyze and track animals. How have the results been disseminated to communities of interest?The ODD was provided and explained to the stakeholder group at the spring 2025 meeting and subsequently revised based upon their feedback. Behavior and movement data was presented through a poster at the annual Poultry Science Associatoin Meeting. What do you plan to do during the next reporting period to accomplish the goals?Finish coding the ABM for the floor system. Then compare its predictions to empirical data from CTMM analysis of chicken movements and revise ABM code as needed to validate the ABM. Thenextrapolate that mode to the tiered experimental system data and validate that version of the model. Then use those models to compare and contrast scenarios for designing and building better cage free housing systems. Finally, write the publications that accompany the above systems and publish those papers in relevant peer reviewedjournals.Continue gathering and analyzing data pertaining to the extension survey.

Impacts
What was accomplished under these goals? For specific Aim 3 the ODD for the ABM was completed (although it is a living document so details are still being added). The ABM code for working versions of all sub models have been completed and will soon be integrated into the full model. For specific Aim 4 Dr. Cain worked extensively with Gideon from Dr. Erasmus' lab to insure that the digital records of the movements of chickens in the experimental system are formatted correctly for statistical analysis to provide quantitative estimates of empirical patterns to use when validating AMB performance.

Publications


    Progress 02/01/23 to 01/31/24

    Outputs
    Target Audience:Researchers, academic audiences and stakeholder advisory group Changes/Problems:Some minor challenges have presented themselves, but nothing that wasn't addressed.Our original plan was to use several blood endocrine and cellular markers to assess relative affective states associated with housing environments and age.However, since submission of the proposal, the Fraley and Karcher labs have learned, and published, that these humoral variables are insufficient at best, and likely inappropriate to determine avian welfare/affective state associated with chronic conditions, such as housing environments.Thus the Fraley lab has moved to a purely neuroendocrine assessment of these states as described above. Salmonellawas detected in the cage-free housing systems and flocks, PIs have been notified to determine how Salmonella was introduced and what improvements can be made to biosecurity. What opportunities for training and professional development has the project provided?Dr. Fraley has been working with his Masters student, Melanie Bergman, to prepare her for a career in the poultry industry.This training has involved learning to design experiments, relate physiology and biochemical measures to management decisions and production.Further, Melanie has learned to analyze and interpret her data, and most importantly, to present it to scientists, stakeholders, and farmers.Melanie is also learning to be a leader as she supervises a team of undergraduates who assist her with her lab responsibilities as well as to develop tangentially related projects that the undergrads work on. Using funding from the project, Dr. Cain was able to attend a workshop while attending The Wildlife Society 2023 Annual Meeting in November. The course introduced Dr. Cain to several R packages and statistical tools for analyzing and interpreting animal movement data. Quantifying animal movements using the procedures and tools taught in the workshop will be extremely useful for predicting, simulating, and understanding hen movements in our virtual systems. This approach will be novel to the animal sciences and provide critical information need for the pattern-matching process (i.e., making sure we have a validated model that we are comfortable using). This project has provided training opportunities for our student intern, who does not have a background in poultry or in egg production. This project has also given our post-doc the opportunity to manage this collaborative research project. The project has provided opportunities to train a graduate student in experimental design, data collection, data analysis, scientific writing, and supervision and mentoring of undergraduate students. The graduate student has also received cross-disciplinary training in poultry science, animal welfare science and computer science. In addition, undergraduate students were trained in animal behavior and welfare data collection and using different types of technology to analyze and track animals. How have the results been disseminated to communities of interest?Findings continue to be shared at the annual stakeholder meetings. What do you plan to do during the next reporting period to accomplish the goals?We will complete the qRT-PCR assays to determine post hatch deep brain photoreceptor expression as affected by housing.Further we will complete qRT-PCR analyses on several neuroendocrine moieties indicative of affective state, such as arginine vasotocin, corticotropin releasing hormone, gonadotropin inhibitory hormone, neuropeptidy Y, among others including several housekeeping genes as benchmarks. The data collected will be analyzed and used to determine the baseline production outcome for flocks within both cage-free housing systems. Data collection pertaining to hen behavior and location will be completed, analyzed, and used to develop agent-based models for new cage-free housing systems. We will continue work with Animal Sciences to develop similar protocols for the extraction of critical chicken movement data from the multi-tier aviary system. Once we have that data we will work on analyzing that data to statistically define patterns of how chickens move and behave in each housing system. Data analysis will be completed on the collected visual system data from the hens.These results will help inform and be integrated into the agent based modeling being developed. Dr. Cain will complete the design phase of the ABM process, which will involve working through a lot of documentation. Dr. Cain will have numerous conversations with project Co-PIs to better understand and incorporate their expertise into the model design and documentation. The information gleaned from these meetings will be compiled into a single "living" document that will fully describe the agent-based model and how it works. Once drafted, there will be opportunities for the project PIs to make comments and suggestions about the design of the model as it is recorded in the "living" document. The "living" document will be ready to share with the advisory board for the members to provide additional feedback during the stakeholders meeting in April of 2024. Any suggestions or concerns that are brought up by the advisory board about the model design will need to be addressed in the "living" document before the model or simulations are coded in NetLogo. Once the ABM documentation is finalized Dr. Cain will code a version of this model in Netlogo. She will then use pattern-oriented modeling approaches to compare the performance of the ABM to empirical data and refine model code as needed. After completing any necessary revisions to the model Dr. Cain will implement the experimental scenarios using that tool. She will then run those scenarios and statistically contract their performance and use those outputs to inform the relative performance of a range of practical approached to meet project objectives of reducing the rate of mislaid eggs while increase bird welfare in cage free systems. Finally, Dr. Cain will lead the development of manuscripts describing the above process and submit them to appropriate peer reviewed journals.

    Impacts
    What was accomplished under these goals? Asurvey for industry stakeholders was developed to gather baseline information about laying hen housing systems, management, behavior and welfare.Thebiochemical and molecular analyses to suggest affective state of the birds in different housing conditions at different ages of development are to be run. The last developmental age was collected during the end of November, 2023.Brains were collected from same birds as eye samples.Brains were microdissected to isolate the whole diencephalon and immediately dorsal cortical neural structures.This block of brain tissue was ground up under liquid nitrogen conditions and RNA isolated.Gels have been run on all RNA isolates from every brain to confirm RNA viability.At this point, all samples are stored at -80C until 100% of brain samples have been collected--completed as of December, 2023. We could not run the analytical assays until all samples were collected so they could be run simultaneously during 2024. The visual parameter data on the floor and aviary hens at age 55 weeks and 85 weeks was collected. Students were trained on how to identify and count oil droplet cells from the retinae of the hens. Data was gathered on the sensitivity of the hens' photoreceptors, the transmittance of the ocular media, density and topography of the photoreceptors, and pecten orientation.Eyes were also harvested and preserved in fixative to examine the density and topography of retinal ganglion cells. Egg samples have been processed to determine egg quality. Baseline egg quality (volscan, shell strength, haugh unit, yolk index, vitelline membrane strength, whole egg solids and shell thickness) outcomes were determined from eggs collected from sister layer flocks raised in two commercial cage-free housing systems. The baseline for the presence ofSalmonellawas also determined for the flocks. Processing of egg and fecal samples for the presence ofSalmonellahas been completed. Hen behavior has been video recorded and recordings are currently being analyzed to develop a dataset of individual hen behavior and resource use within the two types of cage-free housing systems. The Erasmus lab completed video annotations of hens' behavior at the first timepoint and continues to annotate behavior of hens at later timepoints. Students are collecting data on hens' locations, with coordinates, so that this information can be provided to the Zollner lab to develop the agent-based models. The graduate student is developing a detailed literature review to summarize previous studies examining hen behavior and methods used to track hens in cage free housing systems. In addition, the Erasmus lab worked with a computer scientist to develop a method for tracking individual hens on video recordings; this method is currently being used to collect data about hens' location within each housing system. Dr. Cain was hired in July 2023 to lead the development and coding of the agent based models. Dr. Cain spent the first six months of the project learning the NetLogo programming language and familiarizing herself with best practices for Agent Based Modeling. Dr. Cain is leading the development and coding of the agent-based models. Since her appointment in July, Dr. Cain has expanded her knowledge and understanding of ABM development and analysis. Dr. Cain also completed the software training and practical introduction to the NetLogo modeling environment. She is in the midst of developing the ODD documentation as part of the design of the Agent Based Model and has completed a first round of interviews with project Co-PIs to inform model design. Working withGideon Ajibola and Dr. Erasmuswe have confirmed that the system they have developed for video recording systematic samples of chickens in the floor system can provide the data we need to parameterize and evaluate the Agent Based Model.

    Publications


      Progress 02/01/22 to 01/31/23

      Outputs
      Target Audience:Researchers and academic audiences Changes/Problems: Nothing Reported What opportunities for training and professional development has the project provided?The project has provided an opportunity to train four master's students and a PhD. 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? Data collection will continue through 2023. The team will continue analyzing video recordings to decode hen behavior in the different housing systems. The team will complete development of detailed maps of the housing systems so that data can be collected to inform the agent based models. Production and physiological data will continue to be collected and analyzed. The Advisory Committee will be meeting with the research team to discuss progress and discuss surveys to be distributed to the industry.

      Impacts
      What was accomplished under these goals? Collecting data to understand the implications of two styles on cage-free laying housing onSalmonellaspp. prevalence in the flocks,Salmonellaspp. prevalence associated with egg produced, and physical egg quality parameters. The collection ofbaseline information on hen behavior to inform the agent based models (specific aim 3). Parameters needed for visual model development including reflectivity of items within the environment, oil droplet measures from the eyes and eye shape have been collected.

      Publications