Disease on the Wing: Mapping Avian Influenza During My Internship
I’m Atharva Kaprekar — a Master of Preventive Veterinary Medicine student at UC Davis School of Veterinary Medicine and a veterinarian from India. I am someone who gets excited about a combination of wildlife health (particularly birds), messy spreadsheets, noisy ecological data, and turning maps into meaningful public-health stories. I came to the MPVM program because I wanted to bridge wildlife ecology and disease surveillance: to use spatial tools and statistical thinking to answer practical One Health questions. My interests centre on spatial epidemiology, reproducible data pipelines, and translating technical results into insights that can help surveillance and response.
Before the internship itself, I had the opportunity to attend the Ecology and Evolution of Infectious Diseases (EEID) Conference 2025 at the University of Notre Dame, along with its pre-conference workshop. The workshop led by instructors Pranav Pandit, Pranav Kulkarni and Yingying Wang from UC Davis and Valya Kuskova and Nuno Moniz from University of Notre Dame focused on training students in both foundational and emerging quantitative tools for disease ecology. The workshop was led by a mix of academic and non-academic scientists with deep experience translating research into applied contexts, which made the learning feel immediately relevant. We were able to choose between two tracks — one on modelling vector and host abundance and time-dependent data, and another on AI applications in disease ecology and evolution. Through group-based projects, I gained practical experience applying these skills while also engaging in discussions and case studies that connected theory to real-world problems in public health, policy, conservation, management, and decision-making.
Dr. Kaprekar at the 2025 Ecology and Evolution of Infectious Diseases (EEID) Conference at the University of Notre Dame.
For me, this was the perfect transition into the internship. The workshop gave me fresh technical tools, a stronger grasp of modelling approaches, and inspiration from real-world case studies. Walking out of that experience, I felt much more prepared to dive into outbreak data, and it paved the way for my internship to become not just a technical exercise but an applied research journey.
This virtual internship was exactly the hands-on challenge I was looking for. It pushed me beyond coursework into real-world problem-solving cleaning a continental outbreak dataset, extracting environmental signals from gridded climate products, and building the first draft of an analysis pipeline that could be reused and adapted. It was a technical journey, sure — but it was also a personal one: I learned how to cope with setbacks, ask better questions, and celebrate small wins along the way.
This internship was hosted by University College Dublin (UCD) — School of Veterinary Medicine, and I worked closely with a small team and my supervisor Miriam Casey while remaining enrolled in the MPVM program at UC Davis. The experience was a genuinely international collaboration: I learned how surveillance questions and data pipelines are framed in a different policy and research context, and I gained practical experience adapting continental-scale analyses to country-level needs. Midway through the project we deliberately narrowed the focus to the Netherlands, and that decision changed the project in productive ways. Focusing on one country let me move from broad exploratory mapping to a tight, interpretable case study where high-quality poultry density data and dense outbreak records made spatial extraction and hypothesis testing much more manageable. (And yes — I could not resist a tiny joke: I went from UC Davis to UCD. Same three letters, completely different climate and data sources — my inbox never knew what hit it.)
Opening the Door to an Unexpected Project
When I first signed up for a summer internship, I expected some data entry, maybe a literature review, and a couple of meetings. Instead I found myself conducting a small research pipeline: from cleaning an entire continental outbreak dataset to extracting environmental covariates from remote sensing products and building an initial regression model. It was intimidating at first — I’d never imagined I would be the person who could wrangle climate NetCDF files or stitch together thousands of point records across Europe — but that’s what happened. What started as curiosity about how wild bird movements could influence poultry outbreaks turned into a multi-month commitment to making messy real-world data usable and meaningful.
Dr. Kaprekar during his virtual internship hosted by University College Dublin, where he worked closely with Professor Miriam Casey and a small team while remaining enrolled in the MPVM program at UC Davis.
The Project in One Paragraph
I worked with a revised outbreak dataset (cleaned and harmonized), focused on highly pathogenic avian influenza outbreaks across Europe. From there, I integrated a suite of environmental and demographic factors — such as temperature, distance to water bodies, human population density, land cover, and poultry density where available. I also created a simplified bird migration covariate, categorizing species as spring migrants, autumn migrants, wintering, resident, or nomadic. Using these, I explored spatial and temporal links between wild bird and poultry outbreaks through distance-and-time thresholds and quadrant-based mapping. The end goal was a reproducible analysis pipeline and a first draft of a mixed-effects regression model to highlight associations worth further investigation.
The Honest Day-to-Day: What My Work Actually Looked Like
Much of my day-to-day work revolved around turning messy datasets into something coherent. The outbreak files had inconsistent country names, missing coordinates, and many ambiguous host descriptions, so I spent long hours harmonizing formats and reducing the number of “Unknowns.” I also designed a new classification system that grouped hosts into wild, poultry, or captive, with an extra distinction between primary poultry outbreaks (likely wild introductions) and secondary spread between farms. On the environmental side, I taught myself how to handle climate grids to extract temperature data for thousands of outbreak points. I measured distances to rivers and lakes using spatial indexing techniques and streamlined workflows to run efficiently at scale. eBird, another massive dataset, was simplified into broad migration categories so it could contribute meaningfully to the model. To communicate results, I produced seasonal maps, quadrant summaries, and outbreak time series. In the Netherlands case study, overlaying poultry density maps with outbreak locations proved particularly useful for hypothesis testing.
Building the Disease Model
The most exciting and daunting phase was moving from maps to models. After considering different approaches, I developed a mixed-effects regression model. This framework allowed me to test relationships — such as whether outbreaks clustered near water or were more likely in colder months — while still accounting for regional differences. The predictors included temperature, proximity to water, human population density, poultry density, and bird migration categories. Building the model was not without frustrations; sometimes it failed to converge, and other times predictors overlapped. But in working through these challenges, I learned to simplify, refine, and build models in incremental steps. More than just a statistical exercise, the model became a way of framing ecological and epidemiological questions in numbers and letting the data respond.
Findings in Brief
Although preliminary, the results pointed toward some clear and consistent patterns:
- Wild bird outbreaks often preceded poultry cases in nearby areas, suggesting a plausible route of introduction.
- Seasonality was strong, with colder months showing more outbreaks and possible environmental persistence of the virus.
- Clustering occurred in hotspots, shaped by farming intensity and ecological features, rather than being randomly distributed.
- Environmental factors mattered — colder temperatures and proximity to water appeared to increase risk.
- Host classification improved clarity, with primary vs. secondary poultry outbreaks showing different spatial and temporal signatures.
- Migration timing added explanatory power, helping make sense of why outbreaks spiked in certain seasons and regions.
These findings don’t claim to be conclusive, but they highlight promising leads for further confirmatory research.
Where I Struggled — And What Kept Me Going
There were moments when imposter syndrome loomed large, especially when long scripts failed or results didn’t make sense. What kept me going was learning to break complex problems into small, testable steps, documenting everything to keep the process reproducible, and leaning on my supervisor’s feedback whenever I felt stuck. Each hurdle became an opportunity to grow, and small victories — like finally extracting climate data for all outbreaks — kept me motivated.
Technical Skills I Gained
By the end of the internship, I had mastered advanced R workflows for spatial epidemiology, learned how to integrate climate data into epidemiological datasets, and developed confidence in spatial analysis techniques such as distance calculations, overlays, and raster extractions. I also built reproducible pipelines with careful documentation — skills that I now feel confident carrying forward to any public health problem involving messy real-world data.
Memorable Moments (The Human Side)
Some moments stand out vividly. There was the thrill of waking up to a successful overnight run that linked climate data to more than 18,000 outbreak records. There was the pride of printing my first map that revealed seasonal outbreak waves sweeping across Europe — and then sharing it with my cohort. There were the long nights fuelled by instant ramen and tea, and the light-hearted laughter with my supervisor about the confusing overlap between “UC Davis” and “UCD.” These small but memorable experiences made the technical work feel all the more human.
Advice for Future MPVM Students
Looking back, I would encourage future MPVM students to embrace messy data because real insights often lie hidden there. Learn spatial basics early, because tools like coordinate systems and raster will be invaluable. Document everything, since reproducibility is essential, and when adding ecological covariates like migration, start simple before layering in complexity.
What’s Next (and What’s Next for Me)
The next steps include refining the regression model, testing sensitivity to outbreak timing and distance thresholds, and exploring interactions between migration and environmental conditions. I also hope to share portions of the workflow so that future students can build on it. Personally, this internship solidified my commitment to working at the intersection of wildlife, livestock, and human health, and to contributing to surveillance-driven decision-making in my career.
Acknowledgements
I am deeply grateful to everyone who supported this project. A special note of thanks goes to the staff and partners at University College Dublin — School of Veterinary Medicine. Their institutional support, data guidance, and expertise made the project possible. I am especially grateful to Professor Miriam Casey whose mentorship, patience, and practical feedback shaped every stage of the analysis. Her suggestions on host classification and surveillance interpretation directly improved the pipeline. I am also thankful to team which included faculty at University College Dublin and partners throughout Europe- , as well as Ana Periera do Vale, Gerald Barry, Jamie Traltos and Maeve Farrel, Inma Aznar for their timely support and suggestions in every step. Thanks also to the UC Davis faculty and staff, whose constant guidance and encouragement shaped my journey and made this internship possible. And finally, a light-hearted thank you to the irony of UC Davis vs. UCD — two institutions with nearly identical acronyms but very different climates and contexts — for keeping my inbox amusingly confusing all summer.