📄 earth_science

Monitoring sika deer population dynamics under culling on Nakanoshima island, Japan, using thermal infrared UAV survey

This study demonstrates that six-year thermal infrared UAV surveys provide a highly reliable, cost-efficient, and effective method for monitoring long-term sika deer population dynamics and detecting culling-induced declines on Nakanoshima Island, Japan, outperforming traditional ground-based methods in effort while correlating strongly with harvest-based estimates.

Original authors: Yukiko Matsuura, Takashi Ikeda, Takehiko Y Ito, Takuma Watanabe, Daishi Higashide, Hiromasa Igota

Published 2026-07-06

📖 4 min read☕ Coffee break read

CC BY 4.0

Original authors: Yukiko Matsuura, Takashi Ikeda, Takehiko Y Ito, Takuma Watanabe, Daishi Higashide, Hiromasa Igota

Original paper licensed under CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/). ✨ This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer

Imagine a small, isolated island in a Japanese lake, like a private zoo with no fences and no natural predators. For decades, this island was home to a growing herd of sika deer. The problem? The deer were eating the island's plants faster than they could grow back, threatening the ecosystem. To fix this, the government started a "population control" program, which essentially means carefully removing deer to bring the numbers down.

But here's the tricky part: How do you count the deer to know if your plan is working?

Traditionally, researchers had to play a high-stakes game of "hide and seek" on foot. They would send a large team of people (sometimes 30 or more) into the woods to chase the deer toward the lake shore so they could count them. It was exhausting, dangerous (slipping on snow-covered mountains), and took a whole day.

The New Tool: The "Heat-Sniffing" Drone

This paper introduces a high-tech alternative: Thermal Infrared (TIR) Drones.

Think of these drones as flying thermal cameras. Just like a night-vision camera sees heat instead of light, these drones can spot a warm deer hiding in a cold, dark forest. The deer's body heat stands out like a glowing ember against the cold ground, making them impossible to miss, even if they are sitting still or hiding under trees.

The Experiment: A Six-Year Test

The researchers tested this drone method on the island for six years (2019–2024). They flew the drones over the island at night or very early in the morning when the temperature difference between the deer and the ground was sharpest.

They compared the drone counts against two things:

The "Old Way": The traditional ground teams chasing deer.
The "Math Model": A complex computer calculation based on how many deer were removed each year and how many were seen by stationary cameras.

The Results: Faster, Safer, and Accurate

The study found that the drone method was a game-changer for three main reasons:

The "Effort" Savings: The traditional ground team needed about 245 hours of human work (35 people for 7 hours) to get a count. The drone team only needed about 51 hours (a small team flying for a few hours). That means the drone method required only 20% of the effort of the old way. It's like switching from manually digging a garden with a shovel to using a high-speed excavator.
Tracking the Trend: The drone counts matched up well with the math models (a correlation of 0.67). Most importantly, the drones successfully spotted the decline in the deer population. As the culling program worked, the drone numbers went down, proving the method could track changes over time.
Safety and Stealth: The ground teams had to hike through dangerous, snowy terrain. The drone team just walked a short distance to launch the drone. Also, because the drone flew high (about 70–120 meters), the deer barely noticed it. Unlike the ground teams that might scare the deer away or change their behavior, the drones were like silent ghosts, letting researchers see the deer exactly where they naturally were.

The Catch and the Future

There were a few hiccups. In one year (2021), the survey was cancelled because a strong wind crashed the drone into a tree. Also, because the drone survey took two days to cover the whole island, some deer might have moved between the two days, causing the count to be slightly lower than the actual number in a couple of years. The researchers suggest that if they can fly the whole island in a single day, the counts will be even more accurate.

The Bottom Line

This paper claims that using thermal drones is a reliable, cost-effective, and much less labor-intensive way to monitor wildlife populations over the long term. It works particularly well in places like this island where the terrain is tough and the animals are hard to see. While the study focused specifically on sika deer on this one island, the authors believe this "flying heat-sensor" approach could be a standard tool for wildlife managers everywhere who need to count animals without scaring them or exhausting their teams.

Technical Summary: Monitoring Sika Deer Population Dynamics via Thermal UAV Surveys

Problem Statement
Effective wildlife management relies on reliable assessments of animal abundance to set targets, evaluate culling programs, and predict ecological damage. While conventional ground-based monitoring methods (e.g., pedestrian sign counts, direct counts, camera traps) are widely used, they are often labor-intensive, time-consuming, difficult to implement in remote or rugged terrain, and prone to low detection probabilities and observer bias. Although Unmanned Aerial Vehicles (UAVs) equipped with thermal infrared (TIR) cameras have emerged as effective tools for wildlife monitoring, particularly for large terrestrial mammals, most existing studies are short-term. There is a scarcity of research evaluating the efficacy of drone-based TIR surveys over extended timescales (multiple years) under active population management interventions, such as culling.

Methodology
The study was conducted on Nakanoshima Island (approx. 5 km²) in Lake Toya, Hokkaido, Japan, a closed sika deer (Cervus nippon) population subject to a long-term eradication program via culling. The researchers conducted UAV surveys over a six-year period (2019–2024), excluding 2021 due to equipment failure.

Survey Protocol: Surveys were performed during winter or spring (defoliation seasons) to maximize visibility. Flights occurred either in the early morning (within one hour of sunrise) or at night (starting two hours after sunset) to ensure high thermal contrast between the deer and the environment.
Equipment: Two UAV models were utilized: an Inspire 1 V2.0 carrying a Zenmuse XT ZXTB19FP TIR camera (2019–2022) and a Mavic 2 Enterprise Advanced with integrated TIR sensors (2023–2024). Flights were conducted at altitudes of 70–120 m above ground level.
Operational Strategy: Due to the island's topography (three peaks and a ridge) and limited battery life, the area was divided into four flight zones. A team of six (one pilot, one assistant, and four sentries) was required. Sentries were deployed to drive deer out of evergreen coniferous forests (where thermal detection is difficult) prior to the survey and prevent re-entry.
Data Collection: The pilot manually controlled the UAV, identifying heat signatures. When a deer was confirmed, the drone hovered to capture georeferenced photographs. Video recordings were reviewed post-flight to verify counts.
Comparative Analysis: UAV-derived deer counts were compared against Harvest-Based Estimation (HBE) models. The HBE utilized a state-space model integrating:
- Culling numbers (observed data).
- Drive-count data (ground-based counts of deer driven to the lakeshore).
- Camera trap data (abundance indices).
- The model estimated latent variables for population size before ( $N_{Bt}$ ) and after ( $N_{At}$ ) culling, assuming population growth occurs at the start of the fiscal year and declines following culling.

Key Results

Population Trend Detection: The UAV-based estimates successfully detected a declining trend in the sika deer population, consistent with the ongoing culling efforts. The HBE model estimated a 64.8% decrease in population size from 2019 to 2024.
Correlation: There was a positive correlation ( $R = 0.67$ ) between the UAV-based estimates and the HBE estimates.
Efficiency: The UAV surveys required significantly less effort than traditional ground-based drive counts. The average annual UAV survey effort was 51.3 person-hours, whereas the drive-count survey required 245 person-hours (35 participants for 7 hours). Thus, UAV surveys required only 20% of the effort of drive counts.
Detection Capabilities: High-temperature signatures allowed for the identification of deer, as well as tracks and bedding sites. While crows and other birds also exhibited thermal signatures, they were distinguishable from deer based on morphological shape and relative head temperature.
Limitations: In 2020 and 2022, UAV counts were lower than the HBE model's credible interval. The authors attribute this to potential movement of deer between flight zones during the two-day survey window, leading to underestimation.

Significance and Claims
The paper claims that UAV surveys equipped with TIR cameras provide a highly reliable and cost-efficient method for long-term wildlife monitoring compared to traditional ground-based surveys. The study demonstrates that this technology can effectively track population dynamics under active management (culling) over multiple years.

The authors highlight several specific advantages:

Reduced Labor and Safety: UAV surveys minimize the physical risk to personnel (e.g., slipping on snow-covered terrain) and require a smaller team.
Low Disturbance: Flying at 70–120 m altitude resulted in minimal disturbance to the deer, with few observed instances of animals fleeing, ensuring that observed locations represent typical activity ranges.
Scalability for Eradication: As eradication programs progress and deer densities drop, the ability to detect animals over wide areas is critical. UAVs offer a method to locate remaining individuals more efficiently than camera traps (which are point-specific) or drive counts, potentially increasing capture efficiency.
Applicability: The method is presented as applicable to similar environments globally, particularly for monitoring large ungulates in forested or rugged terrains where visual detection is challenging.

The study concludes that while short-term surveys may yield underestimates due to animal movement, completing surveys within a single day could further improve accuracy. The authors suggest that as the population declines toward eradication, the accuracy of these counting methods is expected to improve, making UAVs a vital tool for the final stages of population control.