counter hit xanga

Can Satellite Photography Be Used To Monitor Raptor Nests

by -22 views
  • Loading metrics

Detection of two Arctic birds in Greenland and an endangered bird in Korea using RGB and thermal cameras with an unmanned aeriform vehicle (UAV)

  • Mijin Park,
  • Chang-Uk Hyun

Detection of ii Chill birds in Greenland and an endangered bird in Korea using RGB and thermal cameras with an unmanned aerial vehicle (UAV)

  • Won Immature Lee,
  • Mijin Park,
  • Chang-U.k. Hyun


  • Published: September four, 2022


Unmanned aerial vehicles (UAVs), then-called ‘drones’, take been widely used to monitor wild animals. Here, we tested a UAV with red, green, and blue (RGB) and thermal cameras to detect free-living birds in a high Arctic region in North Greenland and in a restricted surface area in the Republic of Korea. Small-scale flocks of molting pink-footed geese (Anser brachyrhynchus) near sea ice and incubating common ringed plovers (Charadrius hiaticula) in the Arctic environs were chosen for the RGB and thermal image studies. From the acquired images, we built mosaicked RGB images and coregistered thermal images, and estimated the animate being shapes. Our results showed that geese were discriminated in both RGB and thermal images with water and bounding main ice backgrounds. Incubating plover bodies were not distinguished in RGB images due to their cryptic coloration, but they were detected in thermal images with cold background areas in the Arctic surroundings. We further conducted a bullheaded survey in a restricted expanse under military control in Korea near the breeding sites of blackness-faced spoonbill (Platalea minor), which is an endangered species. From UAV flights with RGB and thermal cameras operated out of the restricted expanse, we acquired images of white objects in the mudflats and verified that the objects were resting spoonbills by watching the birds. Nosotros suggest that thermal cameras and UAVs can exist applied to monitor animals in extreme environments and in restricted areas and help researchers notice cryptic wader nests.


Contempo technological advances in unmanned aerial vehicles (UAVs) have enabled exploration at fine spatial resolutions in many ecological studies. Using devices equipped with a UAV system, researchers study vegetation dynamics, ecosystem processes and animal population distributions at a large scale [ane]. UAVs provide two chief benefits: efficiency and effectiveness. From an efficiency perspective, UAVs are serviceable in reducing human bias, thus enhancing accurateness [2]. Massive data, such as a series of images, are obtained at a more than reasonable price past UAVs than by manned aircrafts [three]. Manpower reduction is another benefit of UAVs; UAVs are useful for repetitive operations, peculiarly when researchers accept to behave spatiotemporally consistent studies. Regarding their effectiveness, UAVs equipped with multiple sensors satisfy diverse purposes, such as the measurement of radiative heat fluxes over land and river [4] or topographical changes [five]. In add-on to their efficiency and effectiveness, UAVs assure accessibility to dangerous environments and do non crave equipment to exist installed on the basis.

Emphasizing these strengths, UAVs are now widely used in wild fauna research; in terms of habitat types, they are used from forests to polar regions, and with regard to taxa, mammals and birds are the predominant organisms recorded by UAV studies [6]. Early studies focused on testing the availability of UAVs for monitoring [three–9]. Prior to piloting UAVs in applied field application, a blueprint recognition algorithm was adult to count birds using decoys [7]. A similar methodology was performed in the ocean with inflatable kayaks, representing whale-like targets [8]. After, UAVs were preferentially practical to detect large, terrestrial animals and employed to investigate aquatic wildlife living in inland waters or the ocean. With contempo technologies of high-resolution photography, researchers have adopted this device to monitor bird nesting at a distance, such as by counting the distribution of the blackness-headed dupe (Chroicocephalus ridibundus) nests [9]. Moreover, by piloting UAVs higher up the canopy, Junda et al. [10] could ascertain all nest contents in the area and determine the species of the eggs belonged to, the clutch size, and the number of nestlings.

The awarding of UAVs in polar regions is intensely predictable, especially in glaciology [xi]. There have been limitations to monitoring in polar regions because of the lack of accessibility to bounding main ice, but the use of UAVs may bandage light on this thing. Considering UAVs tin comprehend several foursquare kilometers depending on their size [i] and aviate back to the operators, UAVs are well suited for investigating polar regions [12–14]. In particular, there are many geese molting on body of water ice during the summer season. Because molting geese are not able to fly, they assemble together and float near or on ice floes as an adaptation confronting the take chances of predation. Thus, it is difficult for researchers to monitor them at a close distance due to their sensitivity during molt. Furthermore, many wader species also visit the Arctic for convenance. Most waders are cryptic when incubating nests. Waders accept plumage with similar colors and patterns of their groundwork [fifteen]. Some species hunker motionlessly and protect their nests under predation take chances [16]. Therefore, nesting birds are very well concealed, which hinders the location of wader nests from being determined by researchers.

UAVs have strengths for monitoring endangered species because these species sometimes prefer to live in uninhabited islands or restricted military zones to avert human disturbance [17–nineteen]. Thermal cameras take frequently been employed since the 1960s and have mainly been used to detect nocturnal beliefs [20]. A contempo lapwing (Vanellus vanellus) case written report showed that a thermal camera and UAV can be applied to observe incubating nests [21]. We expect that a thermal camera tin exist used to find such geese on seawater and cryptic wader bird nests on the basis because this blazon of camera enables researchers to discriminate warm birds from their environments. In thermal images from the Arctic environment, a relatively high temperature body is easily distinguishable compared to the cold-temperature footing. Conversely, hot backgrounds tin exist discriminated from the lower radiations temperature of animals in temperate or tropical areas.

Here, nosotros tested a UAV system with red, green, and blue (RGB) and thermal cameras in a loftier Arctic expanse and in a restricted area. In Northeast Greenland National Park, we tested a UAV detection technique on molting pink-footed geese (Anser brachyrhynchus) and incubating mutual ringed plover (Charadrius hiaticula). To best our knowledge, this is the offset to utilise UAV with RGB and thermal cameras for monitoring molting geese and convenance waders in the loftier Arctic. Additionally, for a bullheaded survey in the restricted area of Incheon, the Democracy of Korea, black-faced spoonbills (Platalea modest), an endangered species, was observed using our UAV flight method.

Materials and methods

Ethical statement

This research has been conducted under permissions from the Greenland authorities, and was granted a let that includes the consideration and approval of monitoring pink-footed geese and common ringed plover with unmanned aerial vehicles (permission no. G16-074, C-17-4, C-18-four-09). The local ethics commission (Ministry of Industry and Mineral Resources, the Government of Greenland) specifically reviewed and approved the application for survey license on the utilise of biological resources for commercial and research purposes (in the following: Human action on Biological Resources). Additionally, flights in the Republic of Korea were approbated according to the Sectionalization of Intelligence Information of the Republic of Korea Regular army (permission no. 1380). The determination of the exact study location was made in consultation with the Incheon Metropolitan City government and the local environmental system to avoid disturbing the breeding sites of black-faced spoonbills. All applicable international, national, and/or institutional guidelines for the care and use of animals were followed.

Study site and populations

On 18 July 2022, we visited Sirius Passet on the eastern shore of J. P. Koch Fjord in N Greenland. This is a far due north Arctic area, which is located at the breadth 82°47.6’Due north and longitude 42°thirteen.7’West (Fig 1). There are big populations of molting pink-footed geese in this surface area [22, 23] and iv wader species, including common ringed plover [23]. Since 2022, the aforementioned expedition team of half-dozen researchers has been annually visiting the written report site and monitoring the breeding status of birds during the breeding season [23]. A daily census was conducted in a survey area of approximately 5 kmii, and the bird nest positions were recorded with a GPS (Geo7x handheld, Trimble GeoExplorer, Sunnyvale, CA, U.s.; Global Navigation Satellite System (GNSS) accuracy of one–100 cm). The number of eggs and hatching dates were recorded. GPS positions were used for mapping the nest sites in this area included in another report on the breeding survey, which had been continued since 2022 [23]. The survey have been performed before the flights. The geese were observed near the seashore when the birds were resting on the sea ice or floating on melted water. Small flocks were observed near the sea. The molting individuals were very sensitive to human approach, so nosotros kept a altitude of approximately over i hundred meter from the flocks in order to not disturb the birds. Common ringed plover nests were institute almost the streams with rocks. 4 breeding nests were found during the field survey in 2022.

On 17 April 2022, we monitored a modest mudflat area of 0.05 km2
along the Yellow Sea coast in Incheon, Republic of Korea. At latitude 37°’N and longitude 126°39.1’Eastward (Fig 1). This expanse is under landfill construction and restricted for military command purposes [24]. This area was chosen because birds are expected to feed near the mud flat, simply no human being arroyo is allowed due to military purposes. Additionally, there is a breeding site of black-faced spoonbills v km n of this surface area (“Songdo” commune) [24]. Thus, we expected to detect spoonbills in the mud flat.

Flying sites and conditions

We controlled a quadcopter drone (Phantom four avant-garde, DJI co.) over molting pink-footed geese and common ringed plovers incubating nests in Sirius Passet, Northward Greenland (Fig 1A). Nosotros approached three flocks of pink-footed geese once on the 18th
of July under sunny weather without stiff air current. UAV flights were conducted nigh 4 pm (in local time) virtually the seashore for pinkish-footed geese. The air temperature was 7.eight°C, and the relative humidity was 72%. Ane incubating common ringed plover was approached at five pm on 16 July. Similarly, the temperature was 8.0°C, and the relative humidity was 72%. The drone weighed 1,380 thou, and its diagonal size excluding the propellers was 350 mm. This device had GPS and inertial measurement unit of measurement (IMU) functions to help democratic flyers move along controlled routes during flights. The drone was equipped with a visible light (RGB) photographic camera (Phantom iv photographic camera with one-inch 20 MP sensor) and an additional external thermal camera (FLIR Vue Pro R, xiii mm lens, 640 × 512 pixels sensor) was mounted on the body of the UAV with nadir looking geometry.

Additionally, for a bullheaded test, nosotros operated 2 quadcopter drones on the 17th
of April, 2022 in Incheon, the Republic of korea (Fig 1B), when the weather was sunny without stiff air current. UAV flights were conducted at approximately 2 pm (in local time), approximately one hour ahead of the low tide. The air temperature was 17°C and the relative humidity was 49%. Due to the mechanical problems in attachment of a thermal camera to the drone with RGB photographic camera, we had to perform two separate flights with two different vehicles: one with RGB photographic camera and the other with the thermal camera. The first approach was conducted by a quadcopter drone (DAYA 550) weighing 1,500 g equipped with an external thermal camera (the same FLIR Vue Pro R that nosotros used in Greenland). So, after approximately x min, the second arroyo was performed by a quadcopter drone (Phantom 4, DJI co.) with an RGB camera (Phantom iv photographic camera with 1-inch 20 MP sensor).

Epitome conquering and processing

Boertmann et al. [16] previously conducted the monitoring of 3 geese species (pinkish-footed geese, Barnacle Goose
Branta leucopsis
and Light-bellied Brent Goose
Branta bernicla hrota), simply only pinkish-footed geese were counted in our study expanse (8517 individuals). Lee [23] performed conventional monitoring on land and observed pink-footed geese individuals in the molting stage. In this study, we also detected pinkish-footed geese flocks at the seashore by two observers during the daily survey. The molting individuals did non allow humans to approach closely and avoided the approaching humans past moving to the open up h2o [23]. For the drone images of pink-footed geese, we chose iii flocks at the seashore (xvi, 16, and 5 individuals), which had been previously detected by binoculars (Zeiss Victory FL, x×42). Previous studies indicated that seabirds evidence species and status-specific behavioral response [14, 25, 26], simply over 100 m flights were recommended [26] for approaching resting private birds. Thus, nosotros conducted UAV flights at 110 chiliad in a higher place ground level (AGL) for twenty min, and the operator was on a hill approximately 500 m away from the birds.

The UAV moved at a speed of five thousand/s, and RGB images were manually taken by a remote command when birds were seen in the command screen. The thermal camera was automatically set to have images every second using the minimum interval of the thermal camera to collect as many thermal images as possible during the UAV flights inside the limited fieldwork period. After the flying, nosotros selected 24 RGB images and 69 thermal images, which covered the geese and the surrounding area; there was enough overlap among images that at least 3 images overlapped per pixel in the target area, preventing gaps between images. The RGB images were and then mosaicked using construction from motion (SfM)-based PhotoScan Pro software (Agisoft LLC, Petrograd, Russian federation). The RGB image mosaicking was performed with the following procedure: (i) epitome alignment, (ii) sparse point detection, (iii) dense point cloud structure, (iv) digital tiptop model (DEM) generation, and (v) image orthorectification and mosaicking (due east.g., [21]). The ground sampling distance (GSD) values of the mosaicked RGB prototype and coregistered thermal image were 4.19 cm and 20.37 cm, respectively.

The thermal images were non mosaicked because of the temperature discrepancy between images caused by the thermal camera’s measurement accuracy, i.e., ± 5°C. Instead, a single thermal image showing the best sharpness was selected and coregistered to the higher resolution RGB mosaic image with advisedly and manually selected tie points. A thermal epitome containing geese was used to compare the temperature of geese with the temperature of surrounding areas (ocean, sea ice and land). Using the ‘Create Random Points’ tool in ArcGIS x.3 (ESRI, Redlands, CA, The states), 120 random points were produced from a thermal epitome. Among the 120 points, 110 points were assigned to one of 3 surroundings (ocean, sea ice, land) and 10 points assigned to the boundaries were discarded.

In a previous study on nesting locations of Black-vented Shearwater (Puffinus opisthomelas), drones flew down to 25 m AGL, and no strong response was observed [25]. In other seabird studies on breeding nests, Gentoo (Pygoscelis papua) and Adélie penguins (Pygoscelis adeliae) were reported to exhibit strong responses to drones at low altitudes of 10–20 m in Antarctica [xiii], and 11 southern seabird species showed stiff behavioral postures in response to drones at 10 thou distance in a sub-Antarctic region [14]. Here we tested UAV flights no less than xx m height over the incubating plover. Nosotros conducted stepwise flights from 100 k downwards to 20 thou with a 10 m interval, staying at each height for 1 min and observing the response of the incubating plover. We did non notice any meaning behavioral responses of the plover during the flights. To obtain the common ringed plover images, we flew the drone at 20 yard AGL for v min; the operator was in a tent 100 thousand away from the birds. The flight tiptop for the nesting birds was lower (20 yard) than that for the pinkish-footed geese (110 m) in the open water. After flight, we selected 5 high-quality clear and not blurred RGB images with sufficient overlap between images, as well as a single thermal image, equally we did with the pink-footed geese images. Then, the RGB images were mosaicked in PhotoScan Pro, and the thermal image was co-registered to the mosaicked RGB image. The GSD of the mosaicked RGB images and coregistered thermal images were 0.95 cm and 3.68 cm, respectively. The thermal image was used to compare the plover temperature with the temperature of ii surrounding areas (vegetation and rock, and stream). Using the ‘Create Random Points’ tool in ArcGIS 10.3 that we also used for geese thermal images, 100 random points were produced from the thermal image. Amidst the 100 points, 99 points were assigned to one of two surround (vegetation and rock, and stream), and ane point was discarded because it was in a boundary. Additionally, iii linear transect lines intersecting the nest and other surface types were drawn on the thermal image and were investigated by comparing the nest temperature with other environmental temperature profiles. For the bullheaded survey, we operated UAV flights at 110 m height for 15 min and in a military restricted area with mudflats (approximately 0.05 km2) in Incheon, Korea; the operator was on a hill about 400m away from the birds. In consultation with the military machine office, nosotros caused permission to perform UAV flights for bird monitoring only and the operator stayed out of the restricted zone. We aimed to conduct a blind UAV flight test to monitor birds in the restricted zone with no previous cognition. Similar to the geese monitoring, we determined the flight altitude to over 100 g in height to avoid possible disturbances. A total of 138 RGB images were acquired along the muddy flat surface area (S1 Fig) and at least 9 images were overlapped in the mudflat surface area of 0.05 km2
(S2 Fig).

We did non notice any behavioral reactions, such as an alerted reaction with flight abroad or making alarm calls, during the drone flight. Initially, we did non know the species and number of birds existing in the mudflats. In the control screen, we observed white objects, so we took RGB and thermal images effectually the objects, such that images overlapped in a similar style as the images in the proceeding two cases (S3 Fig).

The air temperature and humidity at the time of the UAV flights were practical to the thermal analysis using FLIR Tools software (FLIR Systems Inc., Wilsonville, Oregon, U.s.). The birds and the surrounding areas were measured using temperature contours extracted from the selected and coregistered thermal images with a ane degree interval. The temperature contours were delineated as lines that connect locations of equal temperature using ‘Contour’ tool in ArcGIS 10.three. The highest contour temperature values that were able to delineate and separate individual birds amidst the contours were designated every bit thresholds for delineating boundaries of bird pixels (east.k., [27]). The temperatures between the birds and other surface types were compared using the selected contours confining only bird pixels and a random sampling approach for other surface types overall the unabridged paradigm. The longest length of the convex hull of each selected profile was designated every bit the conservative size of private birds leading to exclusion of the pixels mixed with the environment. The temperatures were expressed with box-whisker plots; a median line separated the lower quartile from the upper-quartile boxes, extended from the lowest to the highest value.

The thermal images of the geese were visually investigated by three people, independently, and consistent identifications and counts were carried out and confirmed by the temperature contours. In the case of the plover, none of the three investigators was able to discover the plover using RGB images. Although the plover was represented as a few pixels in the thermal image, the distinctive thermal dissimilarity betwixt the plover and relatively common cold backgrounds enabled the detection of the incubating plover by all iii investigators. In the blind survey, all 3 investigators reported identical locations and counts of blackness-faced spoonbills using mosaicked RGB images, and then counts were confirmed past the thermal epitome.


Monitoring of molting geese flocks

From one UAV flight, a total of 37 birds in three flocks (xvi, 16, and 5 individuals) were observed, and this result was consequent with the numbers counted by binoculars and portable zoom cameras by two regular researchers in the field. No significant reactions were observed during the UAV flying.

In the mosaicked RGB images shown in Figs ii and 3A, grey-colored round shape objects were observed and these were identified past the researchers using binoculars and beverage zoom cameras (Fig 2, upper correct in the left RGB image). The RGB images were manually taken by a remote control when birds were observed in the control screen and the thermal photographic camera was automatically set to take images every second as stated higher up. Although the thermal image was coregistered to the mosaicked RGB paradigm, the unlike strategies for taking images and selecting nonblurred thermal images caused differences in imaging time such that the geese had a different organization later the pocket-sized imaging time gap betwixt the RGB paradigm and thermal image in Fig 2. In the thermal images of geese, the highest radiation temperature from the central part of the bird shapes was xiv.3°C and the lowest temperature of the sea ice in the same prototype was -nine.iii°C (Fig 3B). In the profile images with a 1°C interval, the geese and the adjacent temperatures approximately ranged from i to 7°C, and the neighboring individual birds were not discriminated nether 5°C (Fig 3A–3D). The degree contours at the 5°C threshold indicated that the 5°C contours formed sixteen polygons (Fig 3E and 3F), and this number (n = 16) corresponded to the counts from the RGB image.


Fig ii.

Mosaicked RGB images (left, number of pixels = 69) and thermal images (right number of pixels = 24), which cover the geese and the surrounding area with sufficient overlap betwixt images.
The pink-footed geese on the open water that were detected past binoculars by the researchers are presented (upper correct in the left RGB image).


Fig 3.

UAV images of pink-footed geese at the seashore.

(A) Zoomed RGB mosaic image. (B) Zoomed thermal image. (C) 1°C-interval temperature contours in the thermal image. (D) Detailed view of the temperature contours around geese pixels within the area of the white box in (C). (Due east) Contours using a 5°C temperature threshold for a flock of sixteen geese. (F) Detailed view of the temperature contours with 5°C temperature threshold of the black box in (E) that provided the conservative proximate size of the geese (35 (± 2.5, SE) cm long).

In the thermal epitome (Fig iv), two of the three flocks (xvi and 5 individuals) were detected. During the paradigm conquering, one geese flock moved and was not included in the thermal image. The mean temperatures of the surrounding environments were 0.7°C (body of water), -iii.8°C (bounding main ice) and ix.7°C (land), and the pink-footed geese had a temperature of half dozen.iii°C (Fig 4). The random points assigned to the abovementioned environments were represented by 28, 76, and half-dozen pixels, respectively, with random points assigned to 84 pixels of pink-footed geese in the images (Fig 4). The RGB image and the selected 5°C contour in thermal images of a flock of geese provided the number of geese in the flock, sixteen, and the conservative and proximate size of the geese, which was calculated to be approximately 35 cm (± 2.5, SE) long. Although we tried to confine pixels representing geese just by temperature contouring, some edge pixels were contaminated with thermal radiations from both the backgrounds and the geese. To consider this mixing effect, the extracted sixteen geese polygons were adapted to ±1°C contour polygons. The mean size of geese ranged from xix.i to fifty.6 cm for the ±ane°C contour polygons. In the case of -ane°C adapted contour polygons (i.e., 4°C contour polygons), two pairs of geese polygons were merged into one polygon, and these merged polygons were non considered in the adding of the size of the geese. In the case of +1°C adapted contour polygons (i.e., vi°C contour polygons), one polygon was removed, equally the selected 5°C was the highest temperature depicting an individual goose.


Fig four.

A thermal image including flocks of pink-footed geese and temperature distributions of each group of epitome pixels.

(A) A UAV thermal shot at 110 thousand high of a goose flock (red dots), body of water (blueish circles), sea water ice (grayness circles) and country (light-green circles). (B) Temperature distribution of sixteen pink-footed geese body shapes and randomly selected dots in each group (ocean, sea ice and state) presented in box-whisker plots (a median line separates the lower-quartile from the upper-quartile, and whiskers extend from the lowest to the highest value) of the temperature distribution of the thermal epitome pixels in ocean, sea water ice, state and pinkish-footed goose areas.

Monitoring of an incubating plover

One UAV flight was conducted to record an incubating plover. No behavioral response was observed during the UAV flight.

The plover was non visible in the RGB image (Fig five, left) but the thermal image visualized a candidate location for the nest, which had been previously found by a researcher, with a temperature up to 19.9°C. The temperatures of the incubating plover were distinctively college than those of the surrounding surround (Figs 5 (right) and 6A). The mean temperatures of random points of the environments and the mutual ringed plover were nine.1°C (stream), thirteen.0°C (vegetation and rock) and 18.nine°C (the common ringed plover), and each covered five, 94, and vii pixels, respectively (Fig half dozen). The plover temperature ranged from 17.1 to 19.9°C, and other surrounding temperatures ranged lower than those of plovers (9.8–14.six°C, vegetation and rock; 8.2–10.1°C, stream). The highest temperature amidst the thermal pixels was nineteen.ix°C in the middle of the bird body, and this temperature was at to the lowest degree°C higher than the transect line (Transect A) temperature profile of other land surface types, which ranged from approximately 11.5–14.3°C (Fig 7). Although ii additional transect lines (Transect B and C) passing through other hot spots were analyzed, the transect line with the common ringed plover pixels yielded a prominently higher temperature gap between the peak value and the other values (Transect A, five.6°C) than the other two transect lines (Transect B, 1.half dozen°C; Transect C, ane.7°C).


Fig 5.

A mosaicked RGB paradigm (left) and a coregistered thermal paradigm (right), which cover ane incubating nest (indicated with an arrow).
The incubating plover that had been previously detected by the researchers during the survey is presented (upper correct in the left RGB image).


Fig 6.

A thermal paradigm shot including an incubating plover nest (indicated with an arrow) and the temperature distributions of each image pixel group.

(A) A 20-m-high UAV thermal shot of an incubating plover (ruddy dot), stream (blue circles) and vegetation and rock (green circles). (B) Temperature distribution of i common ringed plover trunk shape and randomly selected dots in each group.


Fig vii.

A thermal image near the plover nest and the surroundings (left) with three transect lines (A in black, B in orange, C in blue; three lines with hot spots) and the transect temperature profiles, i of which includes the incubating plover (Transect A; the pixel in the middle of the bird body was 19.9°C, and parallel lines ranged from 12 to 14°C).

Monitoring of birds in a muddy tidal area

We conducted two separate UAV flights on the same spot and found no behavioral responses of the birds. Along the tidal line, we controlled the UAV and took 138 RGB images at one 2d interval (S1 Fig). With overlapping images, we could build a mosaicked RGB image (S2 Fig). Half-dozen white objects were observed in a pocket-sized flock in the mosaicked RGB image (Fig 8A) and were likewise detected in a coregistered thermal image (Fig 8B and 8C). From the RGB and thermal prototype, we caused the GPS coordinates (S1 and S3 Figs). At the same spot, a flock of black-faced spoonbills was observed by binoculars. By taking photographs of the objects, we could confirm the identification of the birds (Fig 8A, lesser left). In the thermal image, which was taken by the dissever drone flight in 10 min after the RGB photo, the spoonbills had the lowest radiations temperatures (15.4°C) amongst the surroundings (Fig 8B). The mudflat temperature ranged up to 41.9°C under sunny weather. The profile images indicated that an eighteen°C temperature threshold provided countable round shape figures (n = 6) for a flock of spoonbills (Fig 8C). When the thermal epitome was overlaid on the RGB image, five objects were located on the shadow of the spoonbills and one object in the middle was approximately 50 cm away from the bird (Fig 8D).


Fig viii.

A blind survey with UAV images of white objects in a mud flat area.

(A) Zoomed RGB mosaicked image at 110 m acme. (B) Coregistered thermal paradigm at the aforementioned spot with the RGB paradigm. (C) Selected contours using an 18°C temperature threshold on a flock of vi spoonbills provided countable round shape figures. (D) Selected temperature overlay on the zoomed RGB mosaicked image.


Our thermal image results suggest that piloting UAV equipped with RGB and thermal cameras tin be supplementary to conventional ground bird monitoring. In three case studies, nosotros tested the effectiveness of a UAV combined with a thermal photographic camera in an Arctic surround and in a restricted area. The study of the two arctic bird species showed that the numbers and the locations of the birds could be detected with the hotspots of thermal images. In contrast, the radiation temperatures of spoonbills were lower than those of the mudflat; nonetheless, their estimate numbers and locations could however be estimated with thermal images.

Our geese monitoring results suggest that UAVs may exist confluent with traditional monitoring to discover flocks of 10–20 individuals. The thermal images and RGB images were sufficient to determine the circular shapes of geese and count their numbers. Considering the typical goose torso temperature (approximately xl–42°C of the internal body temperature), the radiation temperature of the pink-footed geese was quite low. This may be because the geese were covered with sea water while swimming and partially because of the mixing consequence betwixt the pink-footed geese and the sea h2o around the geese resulting from the big pixel size (20.37 cm). Additionally, animals are rarely observed to maintain the same temperature on their surface and in their cadre in gild not to minimize rut loss [28]. Given the Chill conditions, it is probable that the geese conserved their heat fifty-fifty during molting. Nosotros causeless that the uncertainty range of the temperature in a single thermal paradigm is by and large smaller than the paradigm to prototype uncertainty (± 5°C) due to fundamental mill calibration of the detector array; thus, temperature differences smaller than ± 5°C could be discernable in the thermal paradigm. However, the radiations temperatures (approximately 5–ix°C) did non overlap with the bounding main water temperature (approximately 1–ii°C), so the geese were distinguishable. When we set the threshold at v°C, the geese individual temperature contours did not overlap each other (Fig 3D). The number of geese counted in the thermal image corresponded to the number counted in the RGB image. Thermal cameras can provide images to confirm the presence of the birds using a threshold temperature to dissever birds from the environs.

For the common ringed plover monitoring, the UAV thermal camera provided 19.9°C spots in a 12–14°C transect line. The birds were not easily visible due to the small body size of plovers, merely the temperature pixels of the bird’s radiation were college than those of rocks and vegetation areas. However, nosotros did not exclude the possibility that rocks were warmed enough that their radiation temperature was near the bird radiation temperature. Our study was conducted in articulate sunny weather condition in the afternoon, and then the vegetation and stone expanse showed a mean temperature of xiii.0°C, which was only a 5.9°C different from the hateful temperature of the plovers (mean temperature of the common ringed plover was 18.9°C). The plover temperature region might be caused by the combined outcome of convex topographic features and land surface composition, which was mainly rocks. Limitation in the use of thermal images tin occur when the temperatures of birds and other state surface types are similar to the temperatures of plover and other hot spots on land over diurnal or seasonal country surface temperature variation. The climatological study of land surface temperature (due east.thousand., [29]) before conducting fieldwork can help high-contrast thermal images of loftier contrast exist obtained. Thermal image acquisition under optimal timing with consideration of the thermal inertia, that is, the state surface’south capacity to behave and store heat and radiate information technology outward [thirty], of each land cover type can result in enhanced thermal dissimilarity among various country covers in caused images.

A flock of black-faced spoonbills was detected in a blind survey in the military restricted zone. After taking UAV images, we confirmed the birds past binoculars and portable cameras at ascension tide when the birds walked to a visible place outside of the restricted area. Contrary to our expectation, the radiation temperatures of spoonbills (approximately xv–xix°C) were lower than those of the mud flat expanse (upwards to 41.9°C). In our results, the threshold temperature in the contour image was gear up to 18°C to detect the individual birds. I possibility is that the white body of the birds could increase the reflectance of light, resulting in a lower temperature than the temperature of the surroundings. Some other possibility is that the thermal information could reflect the absurd spots produced by the shadows of the birds. When RGB and thermal data were overlapped with ten min fourth dimension interval, 5 thermal shape of 18°C contour were located in the shadows of the birds and i thermal shape was located about 50 cm away from the bird. Although in that location were no item reactions detected in the sky with the binoculars, this may be due to the slight motility of the mismatched bird, which possibly move during the two drone flights so that the thermal image was in the place where the bird had originally stood in the offset flight. The nighttime spots which look like thermal shadows take different shapes to the birds in the RGB images and in fact look more similar the shadows that the birds are casting in the RGB images.

Black-faced spoonbills are ranked as Endangered (EN) in the IUCN Crimson List, and the total number of adult individuals is expected to be 2,250. Korea is i of their breeding sites. Thus it is very important to monitor Black-faced spoonbills near the breeding sites. UAVs are highlighted every bit conservational tools to detect protected species or to surveil their external conditions. Sumatran orang-utans and Sumatran elephants were detected and counted by conservation drone [17, 19]. Similarly, Mulero-Pázmány et al. [18] tested a rhino anti-poaching system including fixed-wing aircraft. To protect the Western capercaillies (Tetrao urogallus), Weber and Knaus [31] investigated human disturbance such as snowshoe tracks with camera-mounted fixed-wing shipping in the Entlebuch UNESCO Biosphere Reserve. GPS data loggers were attached to a couple of bottom kestrels (Falco naumanni), and their track was imitated past UAV to empathise their mural employ through mosaicked images [32].

As a airplane pilot report to test a UAV-thermal camera system for the detection of bird species, we evaluated thermal images captured using a UAV for efficacy in counting birds on sea water ice and in detecting cryptic bird nests. A single flight of twenty min was enough to cover three flocks 500 meters away from the seashore, and a flight of less than 5 min distinguished one candidate spot for the nest from the surroundings in x square meters that we had previously checked during the survey. The UAV images also provided gauge body size, but the bird lengths from above may not reverberate actual bird sizes considering the images show 2D projections of 3D objects. Nevertheless, in a habitat with low species diversity, it can be useful to distinguish the birds roughly by body length.

In future studies, UAV thermal systems can be used to monitor seabirds in harsh environments and human-restricted areas where researchers are not hands able to detect wader nests from highly cryptic backgrounds. Nonetheless, we exercise not assume that this method tin can be a replacement of conventional surveying methods. Instead, nosotros recollect that this method tin can provide complementary data to distinguish living creatures from environments. For detecting bird nests during incubation, thermal images tin exist used to search for possible nest candidates earlier humans perform a field survey. Also, this written report based on the interpretation by researchers with photo images that RGB and thermal cameras acquired. This method lacks processing images to classify the patterns of the images and identify animals with the distinctive species-specific characteristics such as their shapes and sizes. Thus, the paradigm processing approaches would improve efficiency and accuracy for animal counting and identifying with UAV imagery.

Nosotros consider the upstanding issues of approaching birds using UAVs. A few studies have examined the behavioral responses of wild animals to approaching devices [12, 26, 33] and, can provide relevant guidelines. Amongst these studies, Vas et al. [26] used a small quadcopter, which was similar to our auto, to approach mallards, greenshanks, and flamingos. Although the authors did not find any measurable impacts within 4 m distances, they suggested avoiding vertical approaches and launching the drone more than 100 m from the animals. The geese and spoonbills might have been alerted to the budgeted drones only we did not notice any meaning responses in the drone images of 110 1000 flights. Additionally, we found no particular reactions from the incubating common ringed plover. The plover did non prove whatsoever behavioral responses to the UAV at a height of twenty yard. Because the plover was in the middle of the incubation period, the bird may remain still fifty-fifty if disturbed. However, the geese were in the molting menstruation and seemed to be very sensitive to whatever moving objects. Even if the flying height was greater than 100 m and higher than previous airplane approaches [22], we do not exclude the possibility that the molting geese were affected by the UAV approaches. Thus, we think that it may be necessary to approximate condom approach guidelines by conducting conscientious flying approaches for different bird species considering convenance characteristics.

Supporting information

S1 Fig. Results of image mosaicking using 138 RGB images.

(A) Image locations and mosaicked images. Red dots bespeak the verbal locations of RGB image acquisition during the UAV flight. In the blackness square, white dots were detected. (B) A zoomed image effectually the white dots which were suspected to be birds.



We thank Ji-Hoon Khim for aid in Greenland and Myungho Seo for operating UAVs in Korea. Nosotros as well thank Dr. Tae-Gun Seo for statistic communication. We thank the Villum Inquiry Station and the Station Nord for logistic support. The research was conducted under permission from government of Greenland (permission no. G16-074, C-17-4, C-18-four-09).


  1. ane.
    Anderson K, Gaston KJ. Lightweight unmanned aerial vehicles volition revolutionize spatial ecology. Frontiers in Ecology and the Environment. 2022;xi(iii):138–46.

  2. 2.
    Hodgson JC, Mott R, Baylis SM, Pham TT, Wotherspoon Southward, Kilpatrick AD, et al. Drones count wildlife more accurately and precisely than humans. Methods in Environmental and Evolution. 2022;9(5):1160–7.

  3. iii.
    Watts Air-conditioning, Perry JH, Smith SE, Burgess MA, Wilkinson Be, Szantoi Z, et al. Small Unmanned Aircraft Systems for Depression-Altitude Aerial Surveys. The Journal of Wildlife Management. 2010;74(vii):1614–9.

  4. 4.
    Reineman BD, Lenain Fifty, Statom NM, Melville WK. Development and Testing of Instrumentation for UAV-Based Flux Measurements within Terrestrial and Marine Atmospheric Boundary Layers. Journal of Atmospheric and Oceanic Technology. 2022;30(vii):1295–319.

  5. 5.
    d’Oleire-Oltmanns Due south, Marzolff I, Peter K, Ries J. Unmanned Aerial Vehicle (UAV) for Monitoring Soil Erosion in Morocco. Remote Sensing. 2022;4(11):3390–416.

  6. vi.
    Chabot D, Bird DM. Wildlife enquiry and management methods in the 21st century: Where practise unmanned aircraft fit in? Journal of Unmanned Vehicle Systems. 2022;3(4):137–55.

  7. 7.
    Abd-Elrahman A, Pearlstine Fifty, Percival F. Development of pattern recognition algorithm for automatic bird detection from unmanned aerial vehicle imagery. Surveying and Land Information Science. 2005;65(1):37.

  8. viii.
    Koski WR, Allen T, Ireland D, Buck M, Smith PR, Macrander AM, et al. Evaluation of an unmanned airborne arrangement for monitoring marine mammals. Aquatic Mammals. 2009;35(3):347.

  9. 9.
    SardÀ-Palomera F, Bota G, ViÑOlo C, PallarÉS O, Sazatornil 5, Brotons Fifty, et al. Fine-scale bird monitoring from light unmanned aircraft systems. Ibis. 2022;154(1):177–83.

  10. 10.
    Junda J, Greene E, Bird DM. Proper flight technique for using a small rotary-winged drone shipping to safely, quickly, and accurately survey raptor nests. Journal of Unmanned Vehicle Systems. 2022;iii(4):222–36.

  11. 11.
    Bhardwaj A, Sam L, Akanksha , Martín-Torres FJ, Kumar R. UAVs as remote sensing platform in glaciology: Nowadays applications and futurity prospects. Remote Sensing of Environment. 2022;175:196–204.

  12. 12.
    Rümmler M-C, Mustafa O, Maercker J, Peter H-U, Esefeld J. Measuring the influence of unmanned aerial vehicles on Adélie penguins. Polar Biology. 2022;39(7):1329–34.

  13. xiii.
    Rümmler M-C, Mustafa O, Maercker J, Peter H-U, Esefeld J. Sensitivity of Adélie and Gentoo penguins to various flight activities of a micro UAV. Polar Biology. 2022;41(12):2481–93.

  14. xiv.
    Weimerskirch H, Prudor A, Schull Q. Flights of drones over sub-Antarctic seabirds testify species- and condition-specific behavioural and physiological responses. Polar Biology. 2022;41(2):259–66.

  15. 15.
    Ferns P. Plume colour and pattern in waders. BULLETIN-WADER Written report GROUP. 2003;100:122–9.

  16. xvi.
    Pienkowski MW. Breeding biology and population dynamics of Ringed plovers Charadrius hiaticula in U.k. and Greenland: nest-predation every bit a possible factor limiting distribution and timing of breeding. Journal of Zoology. 2009;202(1):83–114.

  17. 17.
    Koh LP, Wich SA. Dawn of Drone Ecology: Depression-Cost Autonomous Aeriform Vehicles for Conservation. Tropical Conservation Science. 2022;5(2):121–32.

  18. 18.
    Mulero-Pazmany M, Stolper R, van Essen LD, Negro JJ, Sassen T. Remotely piloted aircraft systems as a rhinoceros anti-poaching tool in Africa. PLoS One. 2022;9(one):e83873. Epub 2022/01/15. pmid:24416177; PubMed Central PMCID: PMC3885534.

  19. 19.
    Wich Southward, Dellatore D, Houghton M, Ardi R, Koh LP. A preliminary cess of using conservation drones for Sumatran orang-utan (Pongo abelii) distribution and density. Journal of Unmanned Vehicle Systems. 2022;four(1):45–52.

  20. 20.
    McCafferty DJ, Marsden S. Applications of thermal imaging in avian scientific discipline. Ibis. 2022;155(1):4–15.

  21. 21.
    Israel Thousand, Reinhard A, editors. Detecting nests of lapwing birds with the help of a small unmanned aerial vehicle with thermal photographic camera. Unmanned Aircraft Systems (ICUAS), 2022 International Conference on; 2022: IEEE.

  22. 22.
    Boertmann D, Olsen K, Nielsen RD. Geese in Northeast and North Greenland as recorded on aerial surveys in 2008 and 2009. Dansk Ornitologisk Forenings Tidsskrift. 2022;109(4):206–17.

  23. 23.
    Lee WY. Bird observations in Siriuspasset, Northward Greenland, 2022 and 2022. Dansk Ornitologisk Forenings Tidsskrift. 2022;112(ii):53–8.

  24. 24.
    Ko Y, Schubert DK, Hester RT. A conflict of greens: green development versus habitat preservation–the case of Incheon, Republic of korea. Environment. 2022;53(3):three–17.

  25. 25.
    Albores-Barajas YV, Soldatini C, Ramos-Rodriguez A, Alcala-Santoyo JE, Carmona R, Dell’Omo G. A new use of technology to solve an old problem: Estimating the population size of a burrow nesting seabird. PLoS Ane. 2022;thirteen(9):e0202094. Epub 2022/09/15. pmid:30216342; PubMed Key PMCID: PMC6157828.

  26. 26.
    Vas E, Lescroel A, Duriez O, Boguszewski G, Gremillet D. Approaching birds with drones: first experiments and upstanding guidelines. Biol Lett. 2022;xi(2):20140754. Epub 2022/02/06. pmid:25652220; PubMed Central PMCID: PMC4360097.

  27. 27.
    Lhoest S, Linchant J, Quevauvillers S, Vermeulen C, Lejeune P. How Many Hippos (Homhip): Algorithm for Automatic Counts of Animals with Infra-Carmine Thermal Imagery from Uav. ISPRS—International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences. 2022;Twoscore-3/W3:355–62.

  28. 28.
    McCafferty DJ, Gallon S, Nord A. Challenges of measuring body temperatures of free-ranging birds and mammals. Animal Biotelemetry. 2022;3(1).

  29. 29.
    Burke C, Rashman M, Wich S, Symons A, Theron C, Longmore S. Optimizing observing strategies for monitoring animals using drone-mounted thermal infrared cameras. International Journal of Remote Sensing. 2022;twoscore(2):439–67.

  30. 30.
    Hardgrove C, Moersch J, Whisner S. Thermal imaging of alluvial fans: A new technique for remote classification of sedimentary features. Earth and Planetary Science Letters. 2009;285(1–ii):124–thirty.

  31. 31.
    Weber S, Knaus F. Using drones as a monitoring tool to find show of winter sports activities in a protected mountain area. ecomont (Journal on Protected Mountain Areas Inquiry). 2022;9(1):30–iv.

  32. 32.
    Rodriguez A, Negro JJ, Mulero K, Rodriguez C, Hernandez-Pliego J, Bustamante J. The middle in the sky: combined use of unmanned aeriform systems and GPS data loggers for ecological research and conservation of pocket-size birds. PLoS Ane. 2022;vii(12):e50336. Epub 2022/12/15. pmid:23239979; PubMed Primal PMCID: PMC3519840.

  33. 33.
    Wright MD, Goodman P, Cameron TC. Exploring behavioural responses of shorebirds to impulsive noise. Wildfowl. 2022;threescore(60):150–67.


Posted by: