Pestology Blog

If you’ve spent any time walking around New York City, you know first-hand that rats are everywhere. We know they’re a staple of New York City nightlife and they’ve infiltrated nearly every square inch of New York real estate, but how much do we actually know about what they do when no one is watching? Well, a research team from Basis Research Institute and New York University set out to answer exactly that. Their study, published in July 2025, applied cutting-edge artificial intelligence tools to study rat behavior in subways, sidewalks, and parks in the hopes of giving us one of the most detailed pictures yet of how urban rats move, communicate, and interact. 

To find where to look, the team started with New York City’s 311 complaint hotline data. Reviewing complaint data over a 15-year period, they found that the 311 hotline received roughly 50 reported rat sightings per day with Manhattan topping the charts at around 200 sightings per square mile per year. They also noted that reports spiked in summer months. Armed with this information, the researchers selected complaint-heavy field sites and spent the month of July 2024 recording rats in three Manhattan locations.  

The big challenge with observing rats is that they prefer to forage at night in dark, shadowed areas. To work around this problem, the team used thermal imaging cameras to track the rodents at night when activity was highest. By detecting heat signatures rather than visible light, the thermal cameras could identify and track individual rats even in pitch-black conditions. They were even able to monitor rats partially obscured by objects such as fences and dense vegetation. 

Once they had the footage, the team used an Ai powered real-time object detection model known as YOLO (You Only Look Once) to detect and track individual rats across frames. They also used a sophisticated algorithm known as ByteTrack that allowed them to identify and distinguish rodents from one another, even when rats briefly obstructed behind objects with a 98.1% accuracy rating. From only 10 minutes of video, the researchers extracted 201 individual movement tracking objects and built a probabilistic model to estimate each rat’s true size and 3D position. These results showed that larger rats were observed moving faster than smaller ones, and that moments of coordinated group movement where multiple rats bolted simultaneously were dominated by the bigger animals. If we were to assume here that larger rats were generally older, this would track with behavioral observations that older rodents tend to be more wary of new objects in their surroundings and would be more likely to bolt at the first signs of danger. Whereas younger, dumber rats were less danger averse and more oblivious to perceived threats.  

Beyond tracking movement, the team also wanted to understand the environments where rats operate. Using a smartphone camera, they photographed each foraging site from multiple angles and fed the images into a spatial model that constructs a detailed 3D model of a scene from a sparse set of 2D photos. The models captured and logged environmental details including park vegetation, subway platforms, and city streets. This allowed the team to quantify how sheltered or exposed each environment was, and begin mapping out and quantifying the geometric features of an environment that rats prefer. This information could be used to help PMPs identify and prioritize high-risk areas much more efficiently. 

As if everything they looked at so far wasn’t already enough, the team aimed to address another understudied question: what do wild urban rats actually sound like? Using a wireless ultrasonic microphone called an AudioMoth, the team recorded rat vocalizations across all three field sites during social interactions. These included rodent encounters in subway tunnels, aggressive confrontations near trash cans, and one instance of a single rat foraging inside a garbage bag on a busy sidewalk. As it turns out, not all rodent vocalizations are made in the range that humans can detect. Ultrasonic vocalizations, or USVs, are sounds above the range of human hearing that rats use to communicate. Existing rodent communication research traditionally organized these calls into two categories. Low range calls (22 kHz), which are thought to signal fear or alarm, and high range calls (50 kHz), which are associated with positive signals. What the NYC recordings revealed was far more complicated than that. 

When the researchers compared their field recordings against a large historical database of published rat vocalizations, NYC rat calls consistently fell outside the two established ranges. The calls tended to be shorter in duration than expected and not fitting neatly into either the 22 kHz or 50 kHz categories. That lone rat foraging inside the trash bag, for example, produced a long bout of near-22 kHz calls. Prior lab data would have suggested that this low range was associated with fear or alarm and may suppress feeding behavior. However, that clearly wasn’t what the New York team observed. Instead, the solo rat seemed perfectly content while they were enjoying their all-you-can-eat trash bag all to themselves. The authors suggest this may support a theory that 22 kHz calls function as a security signal in unpredictable environments, a kind of ambient alert that has little to do with immediate predator threat. Wild NYC rats also produced human-audible squeaks during aggressive interactions which happens to be something rat vocalization research has largely overlooked in favor of ultrasonic calls. This raises more questions surrounding rodent vocalization such as are squeaks doing communicative work that we haven’t given them credit for? 

For pest management professionals, this research is painting a whole new picture of how rodents behave and interact in their environment. Which is amazing. But, to me, this study is mind blowing not because of the data it’s produced. But who produced it and how they went about it. Ralph Peterson, a neuroscientist, leveraged a sophisticated combination of groundbreaking technologies (not traditionally used in our urban pest control science space) to give is never-before seen glimpses into how rodents behave in groups, how they respond to their environment, how they communicate. What’s even more mind blowing is that it’s just the tip of the iceberg. Most of what Peterson and his team have collected so far is initial data in proof-of-concept studies. I can’t wait to see what else they’re able to learn in the coming years as they expand their research to more areas around the city, collect more data, and incorporate new algorithms as they evolve. In other words, they’re just getting started. 

If you find any of this interesting, be sure to catch Ralph Peterson’s session at PestWorld 2026 in Grapevine!