Estimating relative density of an invasive ungulate in a biodiversity hotspot using drone-based thermal video surveys

85 Estimating relative density of an invasive ungulate in a biodiversity hotspot using drone-based thermal video surveys Julia Witczuk1, Stanisław Pagacz1, Romain Alliod2 1 Museum and Institute of Zoology, Polish Academy of Sciences, Twarda 51/55, 00-818 Warsaw, Poland 2 Ecotone NC SARL, 8 rue du capitaine bois, 98800 Nouméa, New Caledonia (Fr) Corresponding author: Julia Witczuk ([email protected].pl) Copyright: © Julia Witczuk et al. This is an open access article distributed under terms of the Creative Commons Attribution License (Attribution 4.0 International – CC BY 4.0). Research Article Abstract New Caledonia’s unique terrestrial habitats – primarily mountain rainforest and sclerophyll forest – face significant threats from invasive ungulates, particularly Javan deer (Rusa timorensis), introduced in the 19th century and now widespread. Since deer are an important game and food source for local communities, the management strategy aims to mitigate their detrimental impact on ecosystems while maintaining populations at levels that continue to support hunting. To achieve these objectives, effective and reliable population monitoring methods are essential. Unmanned aerial vehicles (drones) equipped with thermal sensors represent a potentially superior alternative to conventional groundbased methods for ungulate inventories, particularly in remote and difficult-to-access areas. In this study, we explored the feasibility of using a rotor-wing drone and a thermal camera to estimate the relative density of Javan deer in two protected areas: Domaine de Déva (7,319 ha) and Parc provincial des Grandes Fougères (8,098 ha). Visual line-of-sight flights were conducted after sunset at altitudes of 80 and 100 m above ground level, with the camera fixed at 40° or 0° angles. In Déva, we surveyed 10 sampling blocks, and in Grandes Fougères, 4 blocks (each ranging from 56 to 92 ha). In the predominantly open savanna area of Déva, the estimated relative population density was 116 deer/km2 (SE = 26.8), with some blocks exceeding 400 deer/km2. In contrast, in the dense canopy rainforest of Grandes Fougères, the relative density was 7 deer/km2 (SE = 2.8), with a local maximum of 18 deer/ km2. Differences in deer counts between consecutive flights over the same blocks (with time gaps of less than 100 minutes) were minor, demonstrating that drone surveys are highly repeatable – an essential quality for a reliable population monitoring program. To detect a 25% population reduction with a statistical power of 0.8, surveying 10 sampling blocks was sufficient in the high deer density savanna area, whereas approximately 28 blocks would be required in the rainforest. The logistics of drone operations were relatively straightforward in the savanna; however, in the rainforest, we encountered practical difficulties, including a limited number of suitable take-off sites and restricted visibility due to the dense canopy cover. Despite these obstacles, the method proved to be an effective and efficient approach for monitoring deer populations in the challenging landscapes of New Caledonia. Key words: Javan deer, New Caledonia, Rusa deer, Rusa timorensis, thermal infrared camera, unmanned aerial vehicle, wildlife monitoring, wildlife survey Academic editor: Sandro Bertolino Received: 3 May 2025 Accepted: 14 September 2025 Published: 10 October 2025 Citation: Witczuk J, Pagacz S, Alliod R (2025) Estimating relative density of an invasive ungulate in a biodiversity hotspot using drone-based thermal video surveys. NeoBiota 103: 85–106. https://doi.org/10.3897/ neobiota.103.157791 NeoBiota 103: 85–106 (2025) DOI: 10.3897/neobiota.103.157791 Advancing research on alien species and biological invasions A peer-reviewed open-access journal NeoBiota 86 NeoBiota 103: 85–106 (2025), DOI: 10.3897/neobiota.103.157791 Julia Witczuk et al.: Thermal drone surveys of invasive Javan deer in New Caledonia Introduction Biodiversity loss is currently one of the primary challenges in nature conservation. It is driven by habitat degradation, climate change, as well as competition and predation by alien invasive species (Vitousek et al. 1997; Lefeuvre 2006; Kingsford et al. 2009; IPBES 2019). The negative impact of non-native animals is particularly pronounced on isolated oceanic islands (Courchamp et al. 2003; Russell and Kueffer 2019). Their unique, predominantly endemic vegetation, which often evolved without pressure from large mammalian herbivores and therefore did not develop adequate defense mechanisms, is especially vulnerable to the presence of human-introduced domesticated or wild ungulates (Bowen and Van Vuren 1997). Moreover, in the absence of natural predators, herbivore populations can expand rapidly, posing an increasing threat to island ecosystems. Consequently, controlling invasive herbivores is a crucial conservation priority on islands (Courchamp et al. 2003). New Caledonia is a Pacific archipelago located about 1,500 km east of Australia. It is characterized by exceptional biodiversity and a high number of endemic species. Of the more than 3,300 native vascular plant species found there, 75% are endemic (Morat 1993; Isnard and Jaffré 2024; Munzinger et al. 2025), and the archipelago is considered one of the most important biodiversity hotspots on Earth (Myers 1988; Myers et al. 2000). The most valuable terrestrial habitats of Grande-Terre, the archipelago’s largest island – tropical rainforests (83% of endemic species) and dry sclerophyll forests (57% of endemic species) – are severely threatened by mining, logging, fire, agriculture (cattle grazing), and invasive species (Mittermeier et al. 2004; Isnard and Jaffré 2024). While rainforests still cover much of the hard-to-reach inland areas at higher elevations, the small remnants of the sclerophyll forests that have survived, scattered within savannas along the west coast, occupy only 2% of their original area (Bouchet et al. 1995). Javan deer (Rusa timorensis), also known as Rusa deer, were introduced to GrandeTerre in the 1870s. Due to their high adaptability, as well as the lack of competitors and predators, the species quickly spread throughout the island. Currently, their abundance is estimated at several hundred thousand individuals (Barrière and Fort 2021), with the highest densities recorded along the west coast of the island (de Garine-Wichatitsky et al. 2004). The species is now listed as one of the major priority invasive species (Conservatoire d’espaces naturels de Nouvelle-Calédonie 2017). The highly negative impact of deer on the endemic flora is observed both in the remnants of the sclerophyll forest (Bouchet et al. 1995; Gargominy et al. 1996; de Garine-Wichatitsky and Spaggiari 2008; Mansourian et al. 2018) and in the mountain rainforests (Le Bel et al. 2001; Tron et al. 2024). Deer destroy endemic plant species and impede forest regeneration through browsing, debarking, and trampling. They also contribute to the spread of invasive plant species by actively dispersing their seeds (de Garine-Wichatitsky and Spaggiari 2008). Severe overgrazing causes soil erosion and negatively affects water resources on the island (Tramier et al. 2021). Additionally, local farmers are concerned about the damage caused by deer to pastures and agricultural crops (Barrière and Fort 2021). Although the impact of Javan deer on the island’s ecosystems is undoubtedly negative, they also represent a significant source of game meat for local communities, and deer hunting is an important part of their culture (de Garine-Wichatitsky et al. 2004). For this reason, the species cannot be treated solely as a pest to be eradicated from the island. Instead, it is necessary to manage the population in a way that minimizes its negative effects on biodiversity and ag- 87 NeoBiota 103: 85–106 (2025), DOI: 10.3897/neobiota.103.157791 Julia Witczuk et al.: Thermal drone surveys of invasive Javan deer in New Caledonia riculture while maintaining it at a level that preserves its social significance. Achieving such goals requires effective and reliable methods of population monitoring. The Javan deer population management strategy adopted in New Caledonia focuses control efforts on 10 designated priority areas that are particularly valuable for biodiversity and provide essential ecosystem services, such as those critical to the water cycle (Conservation International and Conservatoire d’espaces naturels de Nouvelle-Calédonie 2016). In three of these areas, regular monitoring and intensive professional deer control have been planned within the framework of the PROTEGE project (Pacific Community 2025). Based on previous experience with deer surveys in New Caledonia (de Garine-Wichatitsky and Saint-Andrieux 2003; Roques-Rogery 2008) and the results from tests conducted under the invasive ungulate control plan launched in 2008 (Barrière and Fort 2021; Fort and Barrière 2021), the strategy recommends monitoring populations using indirect methods that measure deer impact on the forest, along with direct animal observations to estimate abundance indices and identify areas of deer concentration. Due to the nocturnal activity of Javan deer and the limited access to the survey areas, thermal camera observations conducted from unmanned aerial vehicles (drones) are considered the most promising new monitoring method. Previous studies in temperate forests have shown that detecting large mammals in drone-acquired thermal imagery is possible even in relatively dense conifer stands (Witczuk et al. 2018). Tests over captive deer enclosures have further demonstrated that this approach can provide accurate animal counts (Beaver et al. 2020; McMahon et al. 2021; Zabel et al. 2023). The method has also proven reliable under natural conditions, with drone surveys producing consistent results when conducted alongside other deer density estimation methods – such as pellet-group counts and camera trapping (McMahon et al. 2022; Baldwin et al. 2023; Finnegan et al. 2024). Moreover, drone flights in difficult mountainous terrain have been reported to be much more feasible and effective than ground-based observational surveys (Trinh-Dinh et al. 2024). The purpose of this study was to test whether drone-based thermal video surveys could be an effective method for monitoring the Javan deer population in the challenging landscapes of New Caledonia. Following preliminary flights to determine optimal flight parameters, we tested the method in two of the island’s main habitat types – open savanna and dense canopy rainforest – and aimed to establish a methodological foundation for future deer inventories in New Caledonia and similar environments. Methods Study areas The research was conducted on Grande-Terre (16,372 km2), the main island of New Caledonia. The island has a tropical climate with two seasons: a dry season (from mid-April to mid-November) and a rainy season (from mid-November to mid-April) (Fotsing and Dumas 2021). Average monthly temperatures range from 20 to 26 °C, and monthly precipitation varies from 55 mm in October to 271 mm in March (World Bank 2025). The island is sparsely populated, with an average population density of about 15 inhabitants per km2. There are no native mammals on the island except bats. Besides the invasive deer and feral pig (Sus scrofa), the island is also inhabited by domestic cattle, goats, and horses. 88 NeoBiota 103: 85–106 (2025), DOI: 10.3897/neobiota.103.157791 Julia Witczuk et al.: Thermal drone surveys of invasive Javan deer in New Caledonia The choice of the study areas was strongly influenced by COVID-19 pandemic restrictions, as most of the mountain rainforest priority zones designated for deer management were inaccessible due to the lockdown. The first stage of the research involved test flights to establish the optimal flight altitude and camera angle. These were conducted at a private hunting and cattle farm, La Cotonnière, Boulouparis (21°51.23'S, 165°54.65'E). The property (350 ha) features hilly terrain with extensive plains and woodlands. For the test flights, we chose flat, grassy areas with numerous deer. The survey flights to estimate relative deer densities were conducted in two protected areas: Domaine de Déva and Parc provincial des Grandes Fougères (Fig. 1). The survey areas were selected to represent two different ecosystems of the island. Domaine de Déva (21°33.55'S, 165°20.05'E) is a 7,319 ha area located on the west coast of the island. The terrain is hilly, with elevations ranging from 0 to 383m a.s.l. and a mean slope of 12°. The area is dominated by savanna – grasslands and shrub with discontinuous tree cover (Figs 1, 2). Sparse forests include the largest remaining patches of sclerophyll forest in New Caledonia. The Parc provincial des Grandes Fougères (21°36.55'S, 165°45.15'E) is situated approximately in the middle of the island’s central mountain range. The park covers 4,546 ha, but with a 700 m buffer around the park boundaries included, our resulting study area was 8,098 ha. The elevation ranges from 157 to 718 m a.s.l., with a mean slope of 21°. The area is dominated by tropical rainforest, interspersed with patches of shrubs and grassland savannas (Figs 1, 2). Figure 1. Land cover of the two study areas and locations of sampling blocks for Javan deer (Rusa timorensis) thermal drone surveys in New Caledonia: Domaine de Déva (A) and Parc provincial des Grandes Fougères (B); both areas are shown at the same scale. 89 NeoBiota 103: 85–106 (2025), DOI: 10.3897/neobiota.103.157791 Julia Witczuk et al.: Thermal drone surveys of invasive Javan deer in New Caledonia Drone and sensor The flights were conducted using a Matrice 300 RTK quadcopter (DJI, Shenzhen, China). The drone is powered by two lithium polymer batteries (capacity 5,935mAh), enabling a flight endurance of up to 40 minutes. The maximum flight speed is 23 m/s, and the drone is resistant to winds up to 12 m/s. The takeoff weight is 9 kg, including two batteries, a gimbal, a camera, and a parachute. Detailed drone specifications can be found on the manufacturer’s website (https:// enterprise.dji.com/matrice-300). The drone was equipped with a hybrid RGB/thermal infrared camera, the Zenmuse H20T (DJI, Shenzhen, China). The thermal component of the camera featured an uncooled, microbolometer focal plane array detector (640 × 512 pixels, 12 μm pixel pitch, spectral range 8–14 μm) with a 13.5 mm focal length, f/1.0 lens. The camera was mounted on a downward gimbal, with the camera’s horizontal axis aligned perpendicularly to the flight path. Video files (30 Hz) in mp4 format were saved on an SD card; the recorded scenes were also viewed in real time on the drone’s remote controller. Georeferencing data and flight parameters were stored in STR (SubRip Text Subtitle) files. Establishing optimal flight altitude and camera angle The quality of video data acquired by drones strongly depends on the flight altitude above the ground and the angle of the camera. Both of these variables affect the ground sampling distance (GSD, i.e., pixel size) and the size of the camera’s instantaneous field of view (video footprint) (Burke et al. 2019). The lower the flight altitude, the smaller the GSD, improving the ability to distinguish small details in the imagery. However, this comes at the cost of reducing the image footprint size, and thus survey efficiency, as more time is required to cover the same area (Fig. 3). To ensure high detectability and accurate identification of animals in thermal imaging, the thermal signature of an animal should be at least 10 pixels long (Witczuk et al. 2018; Burke et al. 2019). As the average body length of Javan deer is between 130 and 170 cm, the maximum acceptable GSD should be 13 cm. Because of the variability in GSD caused by the camera angle and topography, we applied a conservative rule that the GSD should not exceed 10 cm. Figure 2. Landscapes of the two study areas for Javan deer (Rusa timorensis) thermal drone surveys in New Caledonia: Domaine de Déva (A) and Parc provincial des Grandes Fougères (B). 90 NeoBiota 103: 85–106 (2025), DOI: 10.3897/neobiota.103.157791 Julia Witczuk et al.: Thermal drone surveys of invasive Javan deer in New Caledonia Figure 3. Ground sampling distance (at the center of the camera’s field of view) and footprint area (camera’s instantaneous field of view) for a range of flight altitudes (70–120 m) and camera angles (0–50°). For some combinations of altitudes and angles, the resulting ground sampling distances exceeded the 5–12 cm range and are not shown. A combination of 80 m altitude and a 40° camera angle provides a ground sampling distance of < 10 cm while maximizing the footprint area. To assess the differences in video quality acquired at a range of altitudes (70– 120m) and camera angles (0–50°), we conducted test flights at La Cotonnière farm in May and September 2021 (Fig. 4). Recorded videos were visually inspected, and each altitude–camera angle combination was rated based on the level of detail enabling easy detection and identification of deer while maximizing the footprint area. From this analysis, we established that the optimal parameters for survey flights were 80 m altitude above ground level and a 40° camera angle. These settings result in a GSD resolution of ~9 cm (in the center of the camera’s field of view) and a filmed strip width of about 76 m (calculated based on the formulas provided by Burke et al. 2019). During survey flights over the rainforest, we also collected data using a 0° camera angle (pointing directly downward) to enhance detection of animals obscured by trees. For these flights, altitude was increased to 100 m to maintain the desired GSD of ~9 cm. These alternative settings resulted in a narrower filmed strip width of 59 m. Survey flight operation details At Déva, flights were conducted from 23 to 25 September 2021, and at Grandes Fougères on 4 and 5 October 2021 (Table 1). The survey flights were carried out at night, from sunset (around 7 PM) to approximately 1 AM (NCT), as opportunistic observations indicated that deer become more active around sunset, when they emerge from forests to feed in adjacent grasslands, with activity decreasing after midnight. A similar activity pattern was observed in sambar deer (Rusa unicolor) introduced to Australia (Comte et al. 2022). 91 NeoBiota 103: 85–106 (2025), DOI: 10.3897/neobiota.103.157791 Julia Witczuk et al.: Thermal drone surveys of invasive Javan deer in New Caledonia Figure 4. Video frames showing Javan deer (Rusa timorensis) thermal signatures captured by a drone at different altitudes and camera angles: 80 m, 50°, 44 individuals (A); 80 m, 0°, 52 individuals (B); 120 m, 50°, 132 individuals (C); 120 m, 0°, 79 individuals (D). Insets show close-ups of selected thermal signatures from the top and bottom parts of the images (note the different scales in the top and bottom sections of images recorded with a 50° camera angle). Videos were recorded during test flights at La Cotonnière farm, Boulouparis, New Caledonia, in May and September 2021. At each study area, we selected suitable take-off sites for drone operations that offered vehicle access and clear visibility of the surrounding terrain to ensure flight safety and simplify logistics. Simultaneously, we aimed to choose locations that represented the full range of habitats occurring within the study area. Importantly, site selection was independent of the deer distribution. In the Déva study area, we sampled a total of 10 blocks, which ranged in size from 56 to 92 ha (mean 77 ha). The blocks were reached from five take-off locations. In the Grandes Fougères study area, we sampled four blocks (63–84 ha, mean 70 ha), reached from three take-off sites (Figs 1, 5). The size of the sampling blocks was determined by two constraints: the requirement for visual line-of-sight flights – the drone had to remain within 1,800 m of the operator (as per our derogation) – and the need to retain at least 20% of the drone’s battery after surveying a sampling block to ensure a safe return. Given the variability in take-off site conditions (openness, visibility, and topography), the number, location, 92 NeoBiota 103: 85–106 (2025), DOI: 10.3897/neobiota.103.157791 Julia Witczuk et al.: Thermal drone surveys of invasive Javan deer in New Caledonia Table 1. Details of flights and the number of Javan deer (Rusa timorensis) detected during thermal drone surveys in New Caledonia. Differences in the surveyed area between flight repetitions in the Grandes Fougères are due to varying flight altitudes (80 m and 100 m) and camera angles (40° and 0°) used for each flight. Study area Sampling block Number of flights Date Start times Area surveyed (ha) Detected individuals Déva A1 2 25.09.2021 19:02, 20:38 85, 85 47, 64 A2 2 25.09.2021 19:32, 21:07 84, 84 66, 73 B1 2 24.09.2021 19:04, 21:02 60, 60 44, 56 B2 2†24.09.2021 19:37, 21:29 93, 44 237, 154 B3 2 24.09.2021 20:22, 21:56 85, 85 211, 191 C1 2 25.09.2021 22:25, 22:55 70, 70 121, 120 D1 2 23.09.2021, 24.09.2021 23:27, 0:42 79, 79 17, 19 D2 2 24.09.2021 0:15, 1:11 83, 83 28, 33 E1 1 23.09.2021 19:27 84 77 E2 2 23.09.2021 18:52, 20:14 88, 88 98, 68 Total 811, 678 946, 778 Grandes Fougères F1 2 04.10.2021 19:03, 20:12 57, 45 10, 8 F2 2 04.10.2021 19:36, 20:42 75, 53 2, 2 G1 2 05.10.2021 18:55, 19:23 62, 44 0, 1 H1 2 05.10.2021 0:32, 1:03 58, 48 4, 4 Total 252, 190 16, 15 †Second flight incomplete and flight paths of sampling blocks were planned during fieldwork to maximize coverage and minimize the risk of losing communication with the drone. The flight paths within the sampling blocks had the form of parallel transects (Fig. 5) ranging in length from 0.2 to 1.4 km (mean 1.1 km, SD = 0.19 km). Each block contained 6 to 11 transects. The spacing between non-overlapping filmed strips was about 10 m. The total length of all transects was 106.5 km in Déva and 32.2 km in Grandes Fougères. For most of the sampling blocks (12 out of 14), surveys were repeated twice (Table 1). However, due to technical problems at one of the remaining blocks (Déva B2), the second survey was incomplete, and one block (Déva E1) was surveyed only once. Survey flights over a single block lasted between 18 and 25 minutes, excluding the time required for the drone to travel from the take-off site to the sampling block and back. The interval between the first and second surveys ranged from 8 to 100 minutes (mean 54 minutes). In Déva, both survey repetitions were conducted at an altitude of 80 m with the camera fixed at a 40° angle. In Grandes Fougères, the first survey was also conducted from 80 m at a 40° camera angle. However, since the area was mainly covered by dense canopy forest, the second survey was conducted with the camera fixed at 0° and at an altitude of 100 m to enhance animal detectability. The drone’s speed ranged between 8 and 10 m/s. The survey missions were programmed using Mission Planner software (ArduPilot) and flown autonomously, with only take-off and landing performed manually. Flight parameters and the drone’s location were monitored by the pilot on a ground control station and the screen of the remote controller. Each block was intended to be surveyed using one pair of batteries; however, we assumed the mission would be interrupted and the drone returned to the take-off site if the battery charge dropped to 20%. During all flights, the drone was observed by a second person – the spotter. 93 NeoBiota 103: 85–106 (2025), DOI: 10.3897/neobiota.103.157791 Julia Witczuk et al.: Thermal drone surveys of invasive Javan deer in New Caledonia Video analysis Videos were reviewed manually by two independent observers using the Full Motion Video (FMV) plug-in for QGIS (Raga 2021). This plug-in allows for the analysis and visualization of drone videos within a geographic information system (GIS) environment. FMV requires metadata compatible with the Motion Imagery Standard Board (MISB); however, the STR metadata generated by the H20T camera are not MISB-compatible. To overcome this, we developed a Python script to convert the STR files to the required format. While reviewing files in FMV, when a deer or group of deer was detected, we marked its location, or the center of the group, by placing a point on the video display (signatures for which identification was uncertain were omitted). This action simultaneously created a corresponding point on the map layer, which was later saved in ESRI shapefile format. For each detection, we recorded the timestamp, number of animals, and observed behavior in an Excel spreadsheet. Behavior (generalized at the group level) was classified into one of four categories: 1) running, 2) walking, 3) moving (i.e., remaining in the initial location but changing body position), or 4) still. Figure 5. Distribution of Javan deer (Rusa timorensis) during the thermal drone survey in three sampling blocks (B1, B2, B3; first flight) at the Domaine de Déva study area, New Caledonia. Total number of deer detected: 44 (B1), 237 (B2), 211 (B3). Background image: Gouvernement de la Nouvelle-Calédonie / DITTT / Service topographique. 100 NeoBiota 103: 85–106 (2025), DOI: 10.3897/neobiota.103.157791 Julia Witczuk et al.: Thermal drone surveys of invasive Javan deer in New Caledonia Aerial thermography can be recorded as video files or still images. Our experience from the current study, as well as from previous research (Witczuk et al. 2018; Pagacz and Witczuk 2023), indicates that video data offer significant advantages. The movement of animals captured on video helps distinguish their thermal signatures from other similarly sized hot objects, thereby aiding in detection and recognition. Additionally, video recorded at 30 frames per second – compared to the typical one image per second for still imagery – offers a higher probability of detecting animals, particularly in forested areas, by capturing multiple views of a given location from shifting drone perspectives. It is also important that video footage allows observation of animals’ responses to the drone and the direction of their movement, which – if not random – may lead to biased population estimates. Limitations Drone-based surveys proved to be less effective in dense canopy rainforests than in open areas. In New Caledonia, however, this limitation is less significant, as rainforests cover only about 20% of the island (Allenbach 2021), while shrubs and grasslands account for over 55% (Gouvernement de la Nouvelle-Calédonie and Observatoire de l’Environnement en Nouvelle-Calédonie 2019). To enhance survey effectiveness in mountainous rainforest priority zones, we recommend planning sampling blocks in areas where closed-canopy forest cover does not exceed 50% and allocating greater sampling effort to adjacent savanna areas where deer tend to congregate for foraging. Thermal drone survey limitations also relate to the technical aspects of data analysis. The videos captured during our surveys were manually reviewed using an open-source FMV plug-in for QGIS. This software is basic and lacks several useful features. For instance, while it is possible to create a GIS point layer indicating deer locations by selecting points in the video display window, it is not possible to simultaneously assign attributes such as group size or animal behavior. These attributes must be entered and saved externally in spreadsheet software, which is both time-consuming and prone to errors. The time required to review a video of approximately 40 minutes (recorded over an average-sized sample block of about 70 ha) ranges from 60 to 90 minutes, depending on the number of deer detected and the operator’s experience. The procedure is highly fatiguing, making it difficult to review more than two video files consecutively. The solution would be an automatic video analysis using deep learning algorithms (Sudholz et al. 2021). While commercially available software enabling object detection and tracking in a geographic context is limited and expensive (e.g., the FMV for ArcGIS Pro from ESRI), developing new, original software requires time and the employment of highly qualified personnel. In any case, it is first necessary to create a database for training, testing, and validating the detection/classification algorithm. The data collected during this study provide a good starting point for building such a database and also offer an opportunity to develop a first version of the algorithm, laying the groundwork for future investment in its development. Upcoming deer surveys in New Caledonia are expected to be analyzed using deep learning algorithms, which should significantly reduce the time required for data analysis. The low resolution of commonly used thermal sensors (640 × 512 pixels) is another limiting factor. To achieve an acceptable level of detail in the obtained images, the flight altitude must be relatively low (≤ 100 m), which limits the strip width and reduces the efficiency of the surveys. However, recently introduced high-reso- 101 NeoBiota 103: 85–106 (2025), DOI: 10.3897/neobiota.103.157791 Julia Witczuk et al.: Thermal drone surveys of invasive Javan deer in New Caledonia lution thermal cameras (1280 × 1024) offer the possibility of increasing the flight altitude without the cost of detail loss. Additionally, flying at higher altitudes may help overcome communication issues caused by the terrain, thereby extending the effective range of drone surveys. It is only a matter of time before high-resolution cameras become standard in wildlife research, elevating the quality of drone-acquired thermal imaging to an entirely new level. The final limiting factor is regulations. After years of constant changes, regulations regarding the use of drones are now well established. However, to ensure the safety of all airspace users, the rules for drone operations remain quite restrictive. In most cases, conducting nighttime wildlife surveys will require holding a professional drone pilot license and applying for exemptions that allow night operations at the desired altitude or distance. These regulatory procedures can be cumbersome, depending on the location and the applicant’s experience. Therefore, wildlife managers planning drone surveys should be aware that obtaining the necessary permits can be time-consuming. The high requirements for drone operators, combined with the high cost of the equipment, mean that stakeholders will usually not be able to conduct surveys independently. Instead, they will need to outsource the service to a well-equipped and experienced external contractor. Conclusion Drones equipped with thermal cameras are an effective tool for assessing the relative density of invasive Javan deer in the challenging landscapes of New Caledonia. The time efficiency, verifiability of the data, and highly reproducible survey protocol make drone surveys a much better option than alternative ground-based methods. The straightforward interpretation of the data and the strong demonstrative value of the videos facilitate social acceptance of deer management actions – especially professional deer control – by local stakeholders, such as landowners, residents, tribal customary authorities, and politicians. Further improvements to the method will require the development of software for automatic video analysis and the use of high-resolution thermal cameras. Acknowledgments We are grateful to Patrick Barrière, Marco Zaccaroni, and an anonymous reviewer for their valuable comments and edits on the first draft of the manuscript. This study was carried out as part of the PROTEGE project (Pacific Territories Regional Project for Sustainable Ecosystem Management). We thank the team at the Agence néo-Calédonienne de la Biodiversité (ANCB), responsible for implementing PROTEGE project actions in New Caledonia concerning invasive ungulates, in particular Carole Wéma, Géraldine Bidau, Maxime Lebouteiller, and Patrick Barrière. We also acknowledge the ANCB’s “Invasive Ungulates” technical group, which serves as the PROTEGE project’s steering committee for issues related to invasive deer, feral pigs, and feral goats. Our thanks extend to the Pacific Community (SPC) and the Secretariat of the Pacific Regional Environment Programme (SPREP), the delegatee and co-delegatee of the PROTEGE project, respectively. We also wish to thank Jocelyn Amice, Director of SEM Mwé Ara at the Domaine de Déva; Christophe Schall, Manager of Parc provincial des Grandes Fougères; and Philippe Bourgine, owner of the hunting and cattle farm La Cotonnière. 102 NeoBiota 103: 85–106 (2025), DOI: 10.3897/neobiota.103.157791 Julia Witczuk et al.: Thermal drone surveys of invasive Javan deer in New Caledonia Additional information Conflict of interest The authors have declared that no competing interests exist. Ethical statement No ethical statement was reported. Use of AI No use of AI was reported. Funding The research was conducted under the PROTEGE framework (Pacific Territories Regional Project for Sustainable Ecosystem Management, Theme 4 “Invasive Alien Species,” https://protege.spc.int/ fr) and financed by the European Union under the 11th Regional European Development Fund (EDF), the New Caledonian “Agence Rurale,” and the provincial authorities. Author contributions Julia Witczuk: conceptualization; formal analysis; investigation; methodology; writing – original draft (lead); writing – review and editing, visualization Stansław Pagacz: conceptualization; data curation; formal analysis; investigation; methodology; writing – original draft; writing – review and editing; visualization. Romain Alliod: conceptualization; data curation; funding acquisition; investigation; methodology; resources. Author ORCIDs Julia Witczuk https://orcid.org/0000-0003-0808-1768 Stanisław Pagacz https://orcid.org/0000-0002-0650-9575 Data availability All of the data that support the findings of this study are available in the main text. References Allenbach M (2021) The natural environment of New Caledonia. In: Gravelat C (Ed.) Understanding New Caledonia. University Presses of New Caledonia, Nouméa, New Caledonia, 37–48. https://unc.nc/understanding-new-caledonia/ Baldwin RW, Beaver JT, Messinger M, Muday J, Windsor M, Larsen GD, Silman MR, Anderson TM (2023) Camera trap methods and drone thermal surveillance provide reliable, comparable density estimates of large, free-ranging ungulates. Animals (Basel) 13(11): 1884. https://doi. org/10.3390/ani13111884 Barrière P, Fort C (2021) Rusa timorensis (de Blainville, 1822). In: Savouré-Soubelet A, Arthur C, Aulagnier S, Body G, Callou C, Haffner P, Marchandeau S, Moutou F, Saint-Andrieux C (Eds) Atlas des mammifères sauvages de France, vol 2 Ongulés et Lagomorphes. Muséum National d’Histoire Naturelle, Paris, 96–101. Barry J, Maxwell D (2017) emon: Tools for environmental and ecological survey design. R package version 1.3.2. Comprehensive R Archive Network (CRAN). https://CRAN.R-project.org/package=emon 103 NeoBiota 103: 85–106 (2025), DOI: 10.3897/neobiota.103.157791 Julia Witczuk et al.: Thermal drone surveys of invasive Javan deer in New Caledonia Beaver JT, Baldwin RW, Messinger M, Newbolt CH, Ditchkoff SS, Silman MR (2020) Evaluating the use of drones equipped with thermal sensors as an effective method for estimating wildlife. Wildlife Society Bulletin 44(2): 434–443. https://doi.org/10.1002/wsb.1090 Bouchet P, Jaffre T, Veillon J-M (1995) Plant extinction in New Caledonia: Protection of sclerophyll forests urgently needed. Biodiversity and Conservation 4(4): 415–428. https://doi.org/10.1007/ BF00058425 Bowen L, Van Vuren D (1997) Insular endemic plants lack defenses against herbivores. Conservation Biology: The Journal of the Society for Conservation Biology 11(5): 1249–1254. https://doi. org/10.1046/j.1523-1739.1997.96368.x Burke C, Rashman M, Wich S, Symons A, Theron C, Longmore S (2019) Optimizing observing strategies for monitoring animals using drone-mounted thermal infrared cameras. International Journal of Remote Sensing 40(2): 439–467. https://doi.org/10.1080/01431161.2018.1558372 Comte S, Thomas E, Bengsen AJ, Bennett A, Davis NE, Freney S, Jackson SM, White M, Forsyth DM, Brown D, Taylor A (2022) Seasonal and daily activity of non-native sambar deer in and around high-elevation peatlands, south-eastern. Wildlife Research 49(7): 659–672. https://doi. org/10.1071/WR21147 Conservation International, Conservatoire d’espaces naturels de Nouvelle-Calédonie (2016) Eléments de cadrage pour une stratégie de régulation des cerfs en Nouvelle Calédonie: zones prioritaires, vision, objectifs et ressources nécessaires. Nouvelle-Calédonie, 70 pp. Conservatoire d’espaces naturels de Nouvelle-Calédonie (2017) Stratégie de lutte contre les espèces exotiques envahissantes dans les espaces naturels de Nouvelle-Calédonie. Document cadre. 107 pp. https://especes-envahissantes-outremer.fr/wp-content/uploads/2018/03/strategie-de-lutte-contreles-eee-en-nouvelle-caledonie-2017.pdf Courchamp F, Chapuis JL, Pascal M (2003) Mammal invaders on islands: Impact, control and control impact. Biological Reviews of the Cambridge Philosophical Society 78(3): 347–383. https:// doi.org/10.1017/S1464793102006061 de Garine-Wichatitsky M, Saint-Andrieux C (2003) Faisabilité des méthodes de suivi-évaluation des populations de cerf rusa (Cervus timorensis russa) en Nouvelle-Calédonie. Institut Agronomique néo-Calédonien, Païta, Nouvelle-Calédonie, 42 pp. de Garine-Wichatitsky M, Spaggiari J (2008) Alien plants in native sclerophyll forests of New Caledonia: the role of ungulates. In: Blanfort V, Orapa W (Eds) Ecology, impacts and management of invasive plant species in pastoral areas: Proceedings of the Regional workshop on invasive plant species in pastoral areas, Koné (New Caledonia), November 2003. Secretariat of the Pacific Community, Koné, Nouvelle-Calédonie, 80–84. de Garine-Wichatitsky M, Chardonnet P, de Garine I (2004) Management of introduced game species in New Caledonia: Reconciling biodiversity conservation and resource use? Game and Wildlife Science 21: 697–706. Delplanque A, Linchant J, Vincke X, Lamprey R, Théau J, Vermeulen C, Foucher S, Ouattara A, Kouadio R, Lejeune P (2024) Will artificial intelligence revolutionize aerial surveys? A first largescale semi-automated survey of African wildlife using oblique imagery and deep learning. Ecological Informatics 82: 102679. https://doi.org/10.1016/j.ecoinf.2024.102679 Ditmer MA, Vincent JB, Werden LK, Tanner JC, Laske TG, Iaizzo PA, Garshelis DL, Fieberg JR (2015) Bears show a physiological but limited behavioral response to unmanned aerial vehicles. Current Biology : CB 25(17): 2278–2283. https://doi.org/10.1016/j.cub.2015.07.024 Eikelboom JAJ, Wind J, van de Ven E, Kenana LM, Schroder B, de Knegt HJ, van Langevelde F, Prins HHT (2019) Improving the precision and accuracy of animal population estimates with aerial image object detection. Methods in Ecology and Evolution 10(11): 1875–1887. https:// doi.org/10.1111/2041-210X.13277 104 NeoBiota 103: 85–106 (2025), DOI: 10.3897/neobiota.103.157791 Julia Witczuk et al.: Thermal drone surveys of invasive Javan deer in New Caledonia Finnegan SP, Hinojo A, Monod S, Wall WA, Olsen P, Allen ML (2024) A comparison of thermal drones and camera trap population estimates for Sitka black-tailed deer in Alaska. Wildlife Biology n/a: e01360. https://doi.org/10.1002/wlb3.01360 Fort C, Barrière P (2021) Monographie géographique sur la Nouvelle-Calédonie. In: SavouréSoubelet A, Arthur C, Aulagnier S, Body G, Callou C, Haffner P, Marchandeau S, Moutou F, Saint-Andrieux C (Eds) Atlas des mammifères sauvages de France, vol 2 Ongulés et Lagomorphes. Muséum National d’Histoire Naturelle, Paris, 248–253. Fotsing J-M, Dumas P (2021) The physical and human geography of the New Caledonian archipelago. In: Understanding New Caledonia. University Press of New Caledonia Nouméa, Nouméa, New Caledonia, 19–36. https://unc.nc/understanding-new-caledonia/ Gargominy O, Bouchet P, Pascal M, Jaffre T, Tourneur J-C (1996) Conséquences des introductions d’espèces animales et végétales sur la biodiversité en Nouvelle-Calédonie. Revue d’Écologie 51: 375–402. https://doi.org/10.3406/revec.1996.2217 Gouvernement de la Nouvelle-Calédonie, Observatoire de l’Environnement en Nouvelle-Calédonie (2019) Occupation du sol 2014 au format Shapefile [Dataset]. DINUM / Service de la Géomatique et Télédétection. https://sig-public.gouv.nc/plateforme_telechargement/MOS2014_PN_ PS_PI_SHP.zip [accessed 29.05.2020] Hess SC, Berio Fortini L, Leopold CR, Muise J, Sprague JC (2023) Ecological associations of nonnative ungulates on the Hawaiian Island of Lāna’i. Human-Wildlife Interactions 17: 230–243. https://doi.org/10.26077/be03-b519 IPBES (2019) Global assessment report on biodiversity and ecosystem services of the Intergovernmental Science-Policy Platform on Biodiversity and Ecosystem Services. IPBES secretariat, Bonn, Germany, 56 pp. https://doi.org/10.5281/zenodo.3553579 Isnard S, Jaffré T (2024) What makes New Caledonia’s flora so outstanding? In: Kowasch M, Batterbury SPJ (Eds) Geographies of New Caledonia-Kanaky: Environments, Politics and Cultures. Springer International Publishing, Cham, 21–32. https://doi.org/10.1007/978-3-031-49140-5_3 Kingsford RT, Watson JEM, Lundquist CJ, Venter O, Hughes L, Johnston EL, Atherton J, Gawel M, Keith DA, Mackey BG, Morley C, Possingham HP, Raynor B, Recher HF, Wilson KA (2009) Major conservation policy issues for biodiversity in Oceania. Conservation Biology: The Journal of the Society for Conservation Biology 23(4): 834–840. https://doi.org/10.1111/j.1523-1739.2009.01287.x Le Bel S, Brescia F, Barré N (1999) Etude de la biologie du cerf rusa (Cervus timorensis russa) en milieu naturel, base d’un plan de gestion des populations de cervidés sauvages. Étude de cas: La propriété Metzdorf sur la côte Ouest de la Nouvelle-Calédonie. CIRAD-EMVT, Port-Laguerre, Nouvelle-Caledonie, 63 pp. Le Bel S, Sarrailh JM, Brescia F, Cornu A (2001) Présence du cerf rusa dans le massif de l'Aoupinié en Nouvelle-Calédonie et impact sur le reboisement en kaoris: Gestion de la faune. Bois et Forêts des Tropiques 269: 5–17. https://revues.cirad.fr/index.php/BFT/article/view/20092 Lefeuvre J-C (2006) Les invasions biologiques: un risque pour la biodiversité à l’échelle mondiale. In: Beauvais M-L, Coléno A, Jourdan H (Eds) Les espèces envahissantes dans l’archipel néo-calédonien. IRD Éditions, Paris, CD–ROM: 4–49. https://doi.org/10.4000/books.irdeditions.7656 Mansourian S, Géraux H, Do Khac E, Vallauri D (2018) Lessons learnt from 17 years of restoration in New Caledonia’s dry tropical forest. WWF France, 42 pp. McMahon MC, Ditmer MA, Isaac EJ, Moore SA, Forester JD (2021) Evaluating unmanned aerial systems for the detection and monitoring of moose in northeastern Minnesota. Wildlife Society Bulletin 45(2): 312–324. https://doi.org/10.1002/wsb.1167 McMahon MC, Ditmer MA, Forester JD (2022) Comparing unmanned aerial systems with conventional methodology for surveying a wild white-tailed deer population. Wildlife Research 49(1): 54–65. https://doi.org/10.1071/WR20204 105 NeoBiota 103: 85–106 (2025), DOI: 10.3897/neobiota.103.157791 Julia Witczuk et al.: Thermal drone surveys of invasive Javan deer in New Caledonia Mittermeier R, Gil P, Hoffmann M, Pilgrim J, Brooks T, Mittermeier C, Lamoreux J, Fonseca G (2004) Hotspots Revisited. Earth’s Biologically Richest and Most Endangered Terrestrial Ecoregions. CEMEX, Mexico City, Mexico, 392 pp. Morat P (1993) Our knowledge of the flora of New Caledonia: Endemism and diversity in relation to vegetation types and substrates. Biodiversity Letters 1(3/4): 72–81. https://doi. org/10.2307/2999750 Mulero-Pázmány M, Jenni-Eiermann S, Strebel N, Sattler T, Negro JJ, Tablado Z (2017) Unmanned aircraft systems as a new source of disturbance for wildlife: A systematic review. PLoS One 12(6): e0178448. https://doi.org/10.1371/journal.pone.0178448 Munzinger J, Morat P, Jaffré T, Gâteblé G, Pillon Y, Rouhan G, Bruy D, Veillon J-M, Chalopin M (2025) FLORICAL: checklist of the vascular indigenous flora of New Caledonia. http://publish. plantnet-project.org/project/florical [accessed 31.03.2025] Myers N (1988) Threatened biotas: “Hot spots” in tropical forests. The Environmentalist 8(3): 187– 208. https://doi.org/10.1007/BF02240252 Myers N, Mittermeier RA, Mittermeier CG, da Fonseca GAB, Kent J (2000) Biodiversity hotspots for conservation priorities. Nature 403(6772): 853–858. https://doi.org/10.1038/35002501 Pacific Community (2025) PROTEGE: Pacific Territories Regional Project for Sustainable Ecosystem Management. https://protege.spc.int/en [accessed 01.05.2025] Pagacz S, Witczuk J (2023) Estimating ground surface visibility on thermal images from drone wildlife surveys in forests. Ecological Informatics 78: 102379. https://doi.org/10.1016/j. ecoinf.2023.102379 R Core Team (2025) R: A language and environment for statistical computing. https://www.R-project. org/ [accessed 01.05.2025] Raga F (2021) QGIS Full Motion Video version 1.15. https://github.com/All4Gis/QGISFMV [accessed 01.05.2025] Roques-Rogery G (2008) Monitoring population trends of introduced rusa deer (Cervus timorensis russa) in New Caledonian sclerophyll forests: tests and relevance of methods for management programmes. MS Thesis. Nelson Mandela Metropolitan University (Port Elizabeth, RSA), 88 pp. Russell JC, Kueffer C (2019) Island biodiversity in the Anthropocene. Annual Review of Environment and Resources 44(1): 31–60. https://doi.org/10.1146/annurev-environ-101718-033245 Sudholz A, Denman S, Pople A, Brennan M, Amos M, Hamilton G (2021) A comparison of manual and automated detection of rusa deer (Rusa timorensis) from RPAS-derived thermal imagery. Wildlife Research 49(1): 46–53. https://doi.org/10.1071/WR20169 Tramier CMC, Genthon P, Delvienne QRCP, Sauvan NL, Cassan JJO, Ebrard E, Dumas PS, Queffélean Y (2021) Hydrological regimes in a tropical valley of New Caledonia (SW Pacific): Impacts of wildfires and invasive fauna. Hydrological Processes 35(3): e14071. https://doi.org/10.1002/ hyp.14071 Trinh-Dinh H, Wearn OR, Ngoprasert D, Wich S, Savini T (2024) A drone-based population survey of Delacour’s langur (Trachypithecus delacouri) in the karst forests of northern Vietnam. Biological Conservation 300: 110841. https://doi.org/10.1016/j.biocon.2024.110841 Tron F, Brisset M, Haverkamp C, Barrière R, Aubert M, Theuerkauf J (2024) Exclosure from browsing by invasive ungulates increases species richness and diversity of ground flora in rainforests of New Caledonia. Biological Conservation 296: 110675. https://doi.org/10.1016/j.biocon.2024.110675 Vitousek PM, Mooney HA, Lubchenco J, Melillo JM (1997) Human domination of Earth’s ecosystems. Science 277(5325): 494–499. https://doi.org/10.1126/science.277.5325.494 Witczuk J, Pagacz S, Zmarz A, Cypel M (2018) Exploring the feasibility of unmanned aerial vehicles and thermal imaging for ungulate surveys in forests - preliminary results. International Journal of Remote Sensing 39(15–16): 5504–5521. https://doi.org/10.1080/01431161.2017.1390621 106 NeoBiota 103: 85–106 (2025), DOI: 10.3897/neobiota.103.157791 Julia Witczuk et al.: Thermal drone surveys of invasive Javan deer in New Caledonia World Bank (2025) Current Climate: New Caledonia. Climate Change Knowledge Portal. https:// climateknowledgeportal.worldbank.org/ [Accessed April 11, 2025] Zabel F, Findlay MA, White PJC (2023) Assessment of the accuracy of counting large ungulate species (red deer Cervus elaphus) with UAV-mounted thermal infrared cameras during night flights. Wildlife Biology 2023(3): e01071. https://doi.org/10.1002/wlb3.01071