Remote Sens Ecol Conserv - 2020 - McKellar

You might also like

Download as pdf or txt
Download as pdf or txt
You are on page 1of 13

ORIGINAL RESEARCH

Dual visible-thermal camera approach facilitates drone


surveys of colonial marshbirds
Ann E. McKellar1 , Nicholas G. Shephard1 & Dominique Chabot2
1
Canadian Wildlife Service, Environment and Climate Change Canada, Saskatoon, Canada
2
droneMetrics, Ottawa, Canada

Keywords Abstract
Population census, prairies, remote sensing,
unmanned aerial vehicle, waterbird Waterbirds are important indicators of wetland health, and understanding their
status and trends is necessary for appropriate management and conservation.
Correspondence However, certain species are challenging to survey due to their sensitivity to
Ann E. McKellar, Canadian Wildlife Service, disturbance and the difficulty of accessing their breeding habitats. This is espe-
Environment and Climate Change Canada, cially true for colonially breeding marshbirds, for which a multi-species survey
Saskatoon, SK S7N 0X4, Canada. Tel: 1 306
protocol that maximizes accuracy of counts and minimizes disturbance does
975 4799; Fax: 1 306 975 4089; E-mail
ann.mckellar@canada.ca
not exist. Small drone aircraft have shown promise for conducting accurate and
low-disturbance surveys of colonial waterbirds. However, complex marsh vege-
Editor: Ned Horning tation structures and the cryptic nature of some marshbird nests make them
Associate Editor: Margarita Mulero-Pazmany difficult to detect in aerial imagery. We used synchronous high-resolution visi-
ble imagery (0.8–2 cm/pixel) and thermal-infrared imagery (6–16 cm/pixel)
Received: 10 March 2020; Revised: 21 captured from a drone to count nests of five species of marshbirds at eight
September 2020; Accepted: 30 September
colonies in Saskatchewan, Canada. We compared counts from the imagery to
2020
those obtained from traditional ground-based surveys, generally considered the
doi: 10.1002/rse2.183 most accurate survey method for these species. The two types of imagery
proved highly complementary, as heat signatures helped detect and confirm
Remote Sensing in Ecology and nests not easily spotted in the visible imagery, while the detailed visible imagery
Conservation 2021;7 (2):214–226 allowed species identification. For four species (Western Grebe Aechmophorus
occidentalis, Franklin’s Gull Leucophaeus pipixcan, Forster’s Tern Sterna forsteri,
and Black-crowned Night-Heron Nycticorax nycticorax), drone-based counts
(range 74–1524 nests per colony) were within 5% of ground-based counts
(range 75–1582), while for smaller Black Tern Chlidonias niger colonies (range
3–7 nests), counts from the two methods were always within one nest of each
other. Furthermore, an assessment of flight behavior before, during, and after
drone surveys found no significant evidence of disturbance by the drone. Our
drone-borne dual visible-thermal camera approach proved promising for sur-
veying colonial marshbirds and could potentially be implemented in a range of
situations where wildlife subjects are difficult to survey with visible or thermal
imagery alone.

(e.g., North American Breeding Bird Survey, International


Introduction
Waterbird Census, International Shorebird Survey; Blanco
Monitoring wildlife populations is important for their & Carbonell, 2001; Howe et al., 1989; Sauer et al., 2017),
appropriate management and conservation; in addition, as well as species-specific surveys often aimed at species
certain species can be used as indicators of overall ecosys- of conservation concern (e.g., International Piping Plover
tem health (e.g., wetland birds; Amat & Green, 2010). Census; Elliot-Smith et al., 2015). Such monitoring pro-
Birds are one of the most well-monitored taxa, with grams have proven invaluable in assessing status and
established long-term monitoring programs for certain trends, setting and measuring progress towards popula-
communities of breeding, wintering, and migrating birds tion objectives, and most recently documenting the

214 ª 2020 The Authors. Remote Sensing in Ecology and Conservation published by John Wiley & Sons Ltd on behalf of Zoological Society of London.
This is an open access article under the terms of the Creative Commons Attribution-NonCommercial-NoDerivs License, which permits use and
distribution in any medium, provided the original work is properly cited, the use is non-commercial and no modifications or adaptations are made.
20563485, 2021, 2, Downloaded from https://zslpublications.onlinelibrary.wiley.com/doi/10.1002/rse2.183 by Cochrane Hungary, Wiley Online Library on [30/08/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
A. E. McKellar et al. Drones to Count Marshbirds

devastating declines in North American bird populations minimal (Barr et al., 2020; Chabot et al., 2015; Reintsma
(Butler et al., 2013; Rosenberg et al., 2017, 2019). et al., 2018).
Nonetheless, there remain groups of birds that fall out- An increasingly investigated solution for surveying
side of many of these programs, due to their cryptic nat- wildlife species that are difficult to detect in visible-spec-
ure, unique life histories, and/or difficult-to-access trum drone imagery is the use of thermal-infrared (TIR)
habitats. Colonial waterbirds, and especially those that cameras, which can detect the heat signature of
nest in marshes (hereafter ‘colonial marshbirds’) are one endotherms (e.g., Burke et al., 2019; Chrétien et al., 2015;
such group. And yet, understanding the status and trends Kays et al., 2019; Seymour et al., 2017; Witczuk et al.,
of these species may be especially important for evaluat- 2018). However, the very low spatial resolution of current
ing the health of the sensitive wetland ecosystems they TIR cameras (640 × 512 pixels) makes it challenging to
occupy (Weseloh, 2011). An added complication is that detect smaller animals, such as many bird species, and
many of these species, much like colonial birds in general, to our knowledge there are only two published studies to
can be extremely sensitive to human disturbance; at worst date on the use of drone-borne TIR cameras to survey
they may abandon the colony entirely, or at least they birds (Israel & Reinhard, 2017; Scholten et al., 2019).
may suffer increased predation risk, when disturbed for Although these studies suggest that the heat signatures of
too long or at an inappropriate time of the season (Bur- small birds or their nests may be detectable from up to
ger & Gochfeld, 2009; LaPorte et al., 2013). Several meth- 80 m above ground, the lack of detail in TIR imagery
ods have been developed to survey colonial marshbirds, makes it impossible to confidently identify subjects from
but none is ideal and all suffer important trade-offs the imagery alone in situations where their signatures are
between level of disturbance, accuracy, and required faint compared to the background, where they may be
resources. For example, ground surveys, in which obser- confounded with other objects that radiate heat during
vers enter the colony to count nests, are generally seen as the day (e.g., rocks exposed to the sun), or where it is
the most accurate method but are also the most disrup- necessary to differentiate among multiple species in the
tive given the time required to complete them and the same area. Some studies have addressed this challenge by
presence of observers within the colony. Traditional aerial pairing TIR cameras with higher-resolution visible-
surveys (e.g., fixed-wing or helicopter) may still cause spectrum cameras to aid in identifying mammals
substantial disturbance depending on the altitude flown, (Chrétien et al., 2015; Kays et al., 2019), but this solution
but can generally be completed more quickly; however, has not yet been tested for surveying bird species.
they may be less accurate and may be inappropriate for Here, we report on the use of a dual visible-thermal cam-
smaller species (Hanus et al., 2002; McKellar et al., 2019; era approach for conducting drone surveys of five colonial
Morris, 2006; Steinkamp et al., 2003). marshbird species at eight colonies (including mixed-spe-
Small unmanned aerial vehicles, or drones, may repre- cies colonies) on five lakes in Saskatchewan, Canada. We
sent a convenient, low-disturbance, and accurate method paired drone surveys with traditional ground surveys to
for remotely surveying colonial marshbirds. There has compare the nest counts derived from each method, and
been a major uptake of drone technology in the field of additionally observed flight behavior of the birds over the
wildlife research and management in recent years (Chabot colonies before, during, and after drone surveys to check
& Bird, 2015), and accounts of its use to survey breeding for evidence of disturbance by the drone. We demonstrate
waterbirds have rapidly multiplied (e.g., Chabot et al., the potential of the dual camera approach while making
2015; Hodgson et al., 2016; Lyons et al., 2019; Pöysä recommendations for the improvement of such a system
et al., 2018; Sardà-Palomera et al., 2012). In some cases, for future monitoring of these and other breeding bird
the accuracy of drone surveys has even been shown to populations.
exceed that of traditional ground-based surveys, particu-
larly in situations where the drone can achieve a superior
Materials and Methods
vantage point for observing the birds than ground survey-
ors (Brisson-Curadeau et al., 2017; Hodgson et al., 2018).
Study species
However, the vast majority of published studies on this
subject used visible-spectrum (RGB) cameras, which have The focal colonial marshbird species targeted in our study
been shown to perform more poorly for detecting subjects were Western Grebe Aechmophorus occidentalis, Franklin’s
that contrast weakly with the background (Chabot & Gull Leucophaeus pipixcan, Black Tern Chlidonias niger,
Bird, 2012) as well as those nesting in heavily vegetated Forster’s Tern Sterna forsteri, and Black-crowned Night-
habitats (Barr et al., 2018), both situations that tend to Heron Nycticorax nycticorax. These species have been
apply to marshbirds. Although effects vary by species, for identified as priority species within Bird Conservation
many waterbirds disturbance from drones appears to be Region 11 (BCR 11, Prairie Potholes), where the core of

ª 2020 The Authors. Remote Sensing in Ecology and Conservation published by John Wiley & Sons Ltd on behalf of Zoological Society of London. 215
20563485, 2021, 2, Downloaded from https://zslpublications.onlinelibrary.wiley.com/doi/10.1002/rse2.183 by Cochrane Hungary, Wiley Online Library on [30/08/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
Drones to Count Marshbirds A. E. McKellar et al.

their range occurs in Canada, and Western Grebe is feder- where we initially believed there to be a Franklin’s Gull
ally listed as a species of Special Concern (ECCC, 2013). colony. However, upon performing the ground survey
For the most part, these species are lacking dedicated and examining the drone imagery, we failed to locate gull
regional monitoring programs, with the exception of nests or those of any target species. However, several
Western Grebe which has been surveyed regularly in the American Coot Fulica americana nests were detected, and
province of Alberta (Prescott et al., 2018). these were counted instead, both on the ground and in
In addition to largely overlapping in their Canadian the imagery.
breeding distribution, the species show generally similar
nesting habitat preferences that include shallow freshwater
Ground surveys
marshes with sufficient open water and emergent vegeta-
tion within which nests are built (Burger & Gochfeld, Beginning June 17, we returned to the colonies to con-
2009; Heath et al., 2009; LaPorte et al., 2013; McNicholl duct ground surveys, with the goal of counting all nests
et al., 2001), although nest sites of Black-crowned Night of each target species at each colony. All nest surveys fol-
Herons are more variable and can include stands of live lowed established ground-nest survey protocols for colo-
or dead trees (Hothem et al., 2010). The species often nial marshbirds (Hanus et al., 2002; Jones, 2008;
form breeding colonies on the same waterbodies, even McKellar et al., 2019), and involved two observers wear-
nesting within the same colonies (McKellar et al., 2019). ing chest waders or traveling in kayaks, traversing each
Their nesting phenology largely overlaps within BCR 11, colony in parallel transects spaced 4 m apart and record-
with peak incubation occurring in late May to mid-June ing all nests within 2 m on either side of the transect,
(McKellar et al., 2019), making it possible to conduct until the entire colony was covered. Thus, each observer
multi-species surveys on the same or different waterbodies counted independently the nests within their own transect
for all species concurrently during the nesting season. area, while traversing the transects in parallel to ensure
Breeding colony size can vary substantially both within nests were not double-counted. The exception was at the
and among species, but Franklin’s Gulls generally form Foam Lake Franklin’s Gull colonies (Foam Lake South,
the largest colonies (up to 184 000 individuals; McKellar Foam Lake North), which included in total over 10 000
et al., 2019), whereas Black Terns tend to form smaller nests. It would have been logistically difficult as well as
and looser colonies, occasionally nesting solitarily (Heath extremely time-consuming and disruptive to conduct full
et al., 2009). ground surveys at this site, and instead we performed a
partial ground survey of Foam Lake North. Here two
observers traversed a subset of randomly selected transects
Selection of study sites
(covering approximately 20% of the total colony area)
We searched for nesting colonies throughout late May and recorded all nests within 5 m of the transects. In all
and early June, 2019, targeting waterbodies within BCR cases, observers carried GPS units to ensure they
11 in Saskatchewan where the focal species had previously remained on transect, and distance measurements were
been found to nest (McKellar et al., 2019). Once colonies approximated based on training provided prior to the
with suitable public access or landowner permission to surveys.
the water’s edge were located, observers either waded or During ground surveys, nests that were both active
kayaked the perimeter of the colony while carrying Global (i.e., nests with eggs/chicks observed) and inactive (i.e.,
Positioning System (GPS) units (Garmin GPSMAP 64st, nest appeared intact, but no contents observed) were
Garmin Ltd., Kansas, USA) to delineate the area to be counted for all species except Black Tern. Given the small
subsequently surveyed via ground and drone surveys. For and flimsy nature of Black Tern nests, which tend to
multiple colonies located on the same waterbody, separate degrade quickly when not in use, we did not deem our-
perimeters were taken, and the colonies were given selves capable of reliably counting inactive nests (either
unique names and surveyed separately, when there was an on the ground or in imagery, see below), and so for this
open water break of >100 m between patches of emergent species we only counted active nests. For three of the
vegetation. remaining four species (Western Grebe, Franklin’s Gull,
We located eight colonies hosting at least one of our Forster’s Tern), inactive nests would have only included
focal species on five lakes (Figure 1). We located single- nesting attempts from the current year, given the ten-
species colonies of Western Grebe, Franklin’s Gull, Black dency of marsh nests to degrade over time when not in
Tern, and Black-crowned Night Heron, as well as a use. Black-crowned Night-Heron nests, on the other
multi-species colony that included those first three species hand, are more robust and it is possible that we counted
co-nesting with Forster’s Tern (Jackfish Lake West). One some previous-year nests both on the ground and in ima-
additional study site was located at Old Wives Lake, gery. Indeed, the species often re-uses nests from year to

216 ª 2020 The Authors. Remote Sensing in Ecology and Conservation published by John Wiley & Sons Ltd on behalf of Zoological Society of London.
20563485, 2021, 2, Downloaded from https://zslpublications.onlinelibrary.wiley.com/doi/10.1002/rse2.183 by Cochrane Hungary, Wiley Online Library on [30/08/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
A. E. McKellar et al. Drones to Count Marshbirds

Figure 1. Six lakes in Saskatchewan’s Bird Conservation Region 11 (BCR 11; Prairie Potholes) where drone surveys and associated ground counts
were conducted using a dual visible-thermal camera system at marshbird colonies. Each of the five focal species (clockwise from lower left: Black-
crowned Night-Heron, Forster’s Tern, Western Grebe, Franklin’s Gull, Black Tern) was detected on at least one lake, and a multi-species colony
was found at Jackfish Lake West (Table 1). The inset (lower left) shows the location of BCR 11 in Canada.

year when nesting in trees, as was the case at the Black- Appendix S1 in accordance with the reporting protocol
crowned Night-Heron colonies in our study (Hothem outlined by Barnas et al. (2020). In summary, we con-
et al., 2010). ducted the surveys from June 17 to 21 using a ~9-kg
All ground surveys were performed within 0–7 days of octocopter based on the DJI S1000+ airframe (SZ DJI
the associated drone survey, with one exception. At Quill Technology Co. Ltd., Shenzhen, Guangdong, China) and
Lake South (Black-crowned Night Heron colony), the col- Pixhawk2 flight controller (Hex Technology Ltd., Sha
ony had been completely abandoned sometime between Tin, Hong Kong) carrying a Sony α7R visible camera
the initial site visit in early spring and when we returned (Sony Corporation, Minato, Tokyo, Japan) with a 36-
to conduct ground and drone surveys. Thus, we were able megapixel resolution and 28-mm lens, as well as a FLIR
to conduct the ground survey later in the season (mid- Vue Pro R thermal-infrared camera (FLIR Systems, Wil-
July), as our scheduling allowed, given that Black- sonville, Oregon, USA) with a 640 × 512-pixel resolution
crowned Night Heron nests would not have degraded and 13-mm lens. The drone overflew the entire extent of
over the period of a month. This also meant that only each colony in a grid pattern at an altitude of 45, 60, or
empty nests were counted at this site, both on the ground 120 m above ground (see Appendix S1 for full details)
and in imagery. while the cameras synchronously captured still images,
producing spatial resolutions of 0.8, 1, or 2 cm, respec-
tively, for the visible imagery, and 6, 8, or 16 cm, respec-
Drone surveys
tively, for the TIR imagery. Flight altitudes were selected
Data collection and post-processing with efficiency in mind, generally aiming to fly at the
highest altitude from which we expected to be able to
Full details of the drone surveys, including a description
detect the various species (based on their size and exten-
of the system and its operation as well as image acqui-
sive previous experience performing drone surveys of
sition and post-processing, are provided in the

ª 2020 The Authors. Remote Sensing in Ecology and Conservation published by John Wiley & Sons Ltd on behalf of Zoological Society of London. 217
20563485, 2021, 2, Downloaded from https://zslpublications.onlinelibrary.wiley.com/doi/10.1002/rse2.183 by Cochrane Hungary, Wiley Online Library on [30/08/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
Drones to Count Marshbirds A. E. McKellar et al.

waterbirds) in order to maximize area coverage per flight during three periods: pre-flight, flight, and post-flight.
and in turn minimize the number of successive flights The only exception to this was at Quill Lake North, where
and total time required to survey each colony. any Western Grebes observed off-nest on the water, rather
The individual overlapping images from each camera than in flight, were counted, because this species does not
were subsequently post-processed into orthomosaics (one typically fly except during migration (Hothem et al.,
visible mosaic and one TIR mosaic per colony; Figure 2) 2010). At Jackfish Lake West, where four focal species co-
using OneButton Standard 5.1 (Icaros, Manassas, Vir- nested, only Franklin’s Gulls were recorded for simplicity
ginia, USA), and we then used QGIS 3.2 (QGIS Develop- and because this was the main species observed flying.
ment Team 2018) to manually improve the spatial Observations were made once every 2 min, for up to five
alignment of the paired visible and TIR mosaics. The soft- observations pre and post, and up to 10 observations dur-
ware often failed to mosaic TIR images around the edges ing a flight. Flight observations began as soon as the
of the overflown areas and in areas of open water; how- drone was launched, and ended once it landed. We calcu-
ever, based on our ground observations, these areas did lated the mean number of birds recorded during each
not contain suitable nesting habitat or any nests for all period at each colony, and modeled this as a function of
but one site. At Quill Lake North, which consisted of flight period (pre, flight, post), with colony ID as a ran-
sparse dead trees distributed in featureless open water, the dom effect, using a Gamma distribution. Given that we
software failed to mosaic the TIR images covering almost only flew a single altitude at each colony and at a rela-
all of the surveyed area. Thus, for this colony, our drone- tively small number of colonies, we were not able to eval-
based nest count was based on the visible imagery alone. uate effects of altitude or species composition on
observed disturbance levels (i.e., as interactions in the
model). Nonetheless, with this simple analysis we aimed
Disturbance observations
to evaluate whether, in general, our trials resulted in visi-
During each drone flight, an observer estimated the total bly quantifiable disturbance in the form of birds off their
number of birds that were in flight above the colony nests. We evaluated the significance of flight period using

A B

Figure 2. (A) Visible and (B) thermal-infrared


orthomosaics obtained from a drone flying at
an altitude of 60 m at Foam Lake South.

218 ª 2020 The Authors. Remote Sensing in Ecology and Conservation published by John Wiley & Sons Ltd on behalf of Zoological Society of London.
20563485, 2021, 2, Downloaded from https://zslpublications.onlinelibrary.wiley.com/doi/10.1002/rse2.183 by Cochrane Hungary, Wiley Online Library on [30/08/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
A. E. McKellar et al. Drones to Count Marshbirds

a likelihood ratio test with an alpha level of 0.05. Statisti- (Figure 5). Finally, heat signatures were discernable even
cal analysis was performed in R v. 3.6.3 (R Core Team, in the absence of adult birds or eggs/chicks on the nests,
2020). as was the case at the abandoned Black-crowned Night
Heron colony at Quill Lake South (Figure 6).
Nest counts
Comparison of ground-based and
Prior to counting nests, we created separate point shape-
drone-based nest counts
files in QGIS for each target species at each colony. We
also created a grid to overlay the imagery to help with For all species except Black Tern, nest counts from drone
counting. Size of the grid cells varied by colony and imagery were within 5% of counts from ground surveys
depended on the altitude that the survey was flown and (Table 1). As mentioned above, Black Terns tend to form
the density of nests, and was determined based on the smaller and looser colonies than the other species, so even
preferences of the counter. All imagery was analyzed by a though our ground and imagery counts only differed by a
single individual, who was not involved in the ground single nest at each colony that hosted Black Terns, this
surveys. When counting nests in the imagery, we started amounted to much larger percentage differences between
by examining the visible imagery for a given grid cell. We counts than for the other species (e.g., a single missed
then populated each species-specific shapefile by creating nest resulted in ~28% difference at Jackfish Lake West;
a point at each presumed nest within the visible imagery. Table 1). An exact match between nest counts on the
We then switched the view to the TIR imagery to confirm ground and in the imagery was made for the Black-
what was suspected to be a nest in the visible imagery as crowned Night Heron colony at Porter Lake. In addition,
well as to detect and confirm the location of nests that at Old Wives Lake where we had initially suspected a
were not immediately seen in the visible imagery. This Franklin’s Gull colony to be present (see above), we
often involved switching back and forth between the visi- detected seven American Coot nests both on the ground
ble and TIR views several times. As described above, and in the imagery. We also noted Eared Grebes Podiceps
everything that appeared to be a nest was counted in the nigricollis, another colonial marshbird, co-nesting with
drone imagery regardless of whether a bird or eggs could our focal species at three colonies (Foam Lake South,
be seen on the nest, except for Black Terns for which only Foam Lake North, Jackfish Lake West; Figure S2),
active nests were counted (i.e., bird or eggs visible). All although these nests were not counted.
nests of all target species were then tallied for each colony
where a drone survey was conducted.
Disturbance observations
Given that only a partial ground survey was conducted
at Foam Lake North, we also wished to obtain an imagery The number of birds observed off their nests varied
nest count for the equivalent portion of the colony that among colonies and flight periods (Table S1). At Quill
was surveyed on the ground. To do this, we created a Lake South, there were no Black-crowned Night-Herons
buffer of 5 m radius around each nest counted in the observed during any period, which is not surprising given
imagery and intersected these with the GPS track of the the colony was abandoned. We also did not observe any
surveyed ground transects. We then tallied the number of individuals of that species during the drone trial at Porter
buffered nests that intersected with the transects. Lake, an active colony. These two colonies were excluded
from the statistical analysis. For the remaining colonies,
flight period (pre, flight, post) did not have a statistically
Results
significant effect on the number of birds observed off
their nest (X 22 = 0.088, P = 0.957). Nonetheless, it should
Interpretation of the imagery
be noted that when examining the raw data, at three out
We found nests to be clearly discernable in the visible of six colonies for which birds were observed off their
imagery, as well as their associated heat signatures in the nests, the mean number of birds was greater during the
TIR imagery. For example, at Foam Lake West, overflown flight period than during the pre- or post-flight periods
at an altitude of 45 m, nests of multiple target species (Table S1).
could easily be distinguished, and heat signatures con-
firmed the location of nests (Figures 3 and 4). The above
Discussion
was true even from the highest flight altitude of 120 m at
Foam Lake North, often allowing us to detect nests not In this study, we tested the use of a novel dual camera
easily seen in the visible imagery (Figure S1). We were approach consisting of a high-resolution visible camera
also able to detect the relatively small nests of Black Terns and a thermal-infrared camera mounted on a drone to

ª 2020 The Authors. Remote Sensing in Ecology and Conservation published by John Wiley & Sons Ltd on behalf of Zoological Society of London. 219
20563485, 2021, 2, Downloaded from https://zslpublications.onlinelibrary.wiley.com/doi/10.1002/rse2.183 by Cochrane Hungary, Wiley Online Library on [30/08/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
Drones to Count Marshbirds A. E. McKellar et al.

Figure 3. Four Western Grebe nests (red dots)


and two Franklin’s Gull nests (blue dots) from
(A) visible imagery and (B) thermal-infrared
imagery obtained from a drone flying at an
altitude of 45 m over a multi-species colony at
Jackfish Lake West.

detect and count nests of colonially breeding marshbirds. McKellar et al., 2019; Morris, 2006; Steinkamp et al.,
The two types of imagery proved highly complementary 2003) and some nests in the imagery were likely obscured
for facilitating detection and confirming identification of by overhead vegetation or otherwise difficult to detect.
the often cryptic subjects, allowing us to produce nest However, at the multi-species colony at Jackfish Lake
counts that were very close (within 5%) to counts from West, counts of Western Grebe, Franklin’s Gull, and
traditional ground surveys for four of five species. For Black Tern were higher from the imagery, indicating
Black Terns, percent differences between the two methods some nests were likely missed on the ground or that nests
were higher due to the much smaller size of colonies, but were falsely detected in the imagery (e.g., Barr et al.,
absolute differences were always within a single nest. Our 2018). An additional source of error could be that obser-
results demonstrate the promise of this method for sur- vers on the ground approximated the distance to nests to
veying these and likely other birds breeding in structurally determine which nests to count on their transect.
complex and difficult-to-access habitats, although like any Nonetheless, we observed an exact match between drone
survey method, there are potential sources of error and and ground counts at a Black-crowned Night Heron col-
areas for improvement over our initial approach. ony, as well as during an unplanned survey of American
Coot nests. The latter suggests the potential of our
methodology for other, non-colonial marshbirds.
Comparison to traditional methods
There was no statistically significant difference in num-
For the most part, counts from the drone imagery were ber of birds disturbed before, during, and after our drone
lower than those from the ground. This is to be expected flights. However, at two colonies where Franklin’s Gulls
given that transect-based ground surveys are thought to were counted, as well as at one of the Black Tern colo-
be the most accurate method (Hanus et al., 2002; nies, we observed more birds in flight during the drone

220 ª 2020 The Authors. Remote Sensing in Ecology and Conservation published by John Wiley & Sons Ltd on behalf of Zoological Society of London.
20563485, 2021, 2, Downloaded from https://zslpublications.onlinelibrary.wiley.com/doi/10.1002/rse2.183 by Cochrane Hungary, Wiley Online Library on [30/08/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
A. E. McKellar et al. Drones to Count Marshbirds

A should be noted that birds may still experience physiolog-


ical stress (Weimerskirch et al., 2018). In any case, we
believe disturbance from drone flights is likely still much
less than disturbance from ground surveyors physically
traversing colonies, not only in terms of birds being
scared off nests, but also in total length of disturbance,
disruption to emergent vegetation, and potential attrac-
tion of nest predators (Jones, 2008).

Drone survey parameters and


considerations
At each colony surveyed, we strived to fly the drone at
the highest altitude that would likely allow us to identify
nests of the focal species in the imagery so as to maximize
area coverage per flight, considering that the flight endur-
B ance of multirotor drones powered by industry-standard
lithium-polymer batteries remains limited (in our case,
up to ~20 min). Since the area footprint of individual
images increases with flight altitude, fewer images are
required to cover a given area, translating into fewer and
less tightly spaced flight lines, and ultimately less flight
time. At lower altitudes requiring more images, flight
speed also needs to be reduced to avoid exceeding the
maximum capture rate of the cameras and dropping
images, further amplifying the trade-off between flight
time and altitude.
At Foam Lake North, the largest colony both in terms
of total area (1.219 km2) and number of nests (9598), we
determined that covering the entire colony at our tar-
get altitude for Franklin’s Gulls of 60 m would have taken
at least six successive flights, which we deemed impracti-
Figure 4. Ten Forster’s Tern nests (purple dots) from (A) visible
cal in terms of total time required. We therefore flew the
imagery and (B) thermal-infrared imagery obtained from a drone drone at the highest altitude permitted by Canadian law
flying at an altitude of 45 m over a multi-species colony at Jackfish (120 m), allowing a ~60% increase in the drone’s speed,
Lake West. and resulting in two flights required to cover the colony.
Although Franklin’s Gull nests were detectable from this
altitude, particularly their heat signatures, identification
trial than before or after. This was not the case at three was not always easy and several tens of nests were appar-
other colonies where mean numbers of birds were vari- ently missed, given the higher ground count. At the other
able but never highest during the flight, and no birds end of the spectrum, we flew at an altitude of 45 m over
were observed during any flight period at the active any colonies hosting Black Terns, given their relatively
Black-crowned Night-Heron colony. Given our small small and inconspicuous nests (Heath et al., 2009). To
sample of sites and the fact that flight altitude was con- our knowledge, this is the first aerial survey method to
founded with species composition at each colony, our have successfully detected nests of this species, and it is
disturbance results should be considered preliminary and unlikely that they could be detected from a higher alti-
more detailed studies of potential disturbance by drones tude or with a lower resolution camera. In addition, fly-
for these particular species at various flight altitudes are ing at an altitude of 45 m over the mixed species colony
warranted. Nonetheless, our results are similar to those at Jackfish Lake West may have enabled identification of
from studies of certain waterbirds showing little to no vis- some nests of Western Grebe, Franklin’s Gull, and Black
ible disturbance at similar or lower flight altitudes (Cha- Tern that were not detected on the ground, provided
bot et al., 2015; Reintsma et al., 2018; Magness et al., these were not falsely identified in the imagery. At the
2019; but see Brisson-Curadeau et al., 2017), although it same time, we found that a flight altitude of 60 m over

ª 2020 The Authors. Remote Sensing in Ecology and Conservation published by John Wiley & Sons Ltd on behalf of Zoological Society of London. 221
20563485, 2021, 2, Downloaded from https://zslpublications.onlinelibrary.wiley.com/doi/10.1002/rse2.183 by Cochrane Hungary, Wiley Online Library on [30/08/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
Drones to Count Marshbirds A. E. McKellar et al.

A B

C D

Figure 5. Black Tern nests (yellow dots) at (A) Jackfish Lake East and (B) West, and associated thermal imagery (C and D), obtained from a drone
flying at a height of 45 m. A Western Grebe nest can also be seen to the right of the Black Tern nest in (B and D).

other colonies was sufficient to detect Western Grebe, the background environment, particularly under sunny
Franklin’s Gull, and Black-crowned Night Heron nests. conditions that amplify heating of the latter over the
Ultimately, the detectability of targets in aerial imagery course of the day. One interesting result was the thermal
depends on its spatial resolution, which is a function of detectability of the Black-crowned Night Heron nests on
flight altitude, camera resolution, and lens focal length. Quill Lake South that had been abandoned, indicating
Based on our count results and general assessment of the that the nest materials naturally retain heat more than the
imagery, we would recommend that future drone surveys surroundings, which is not surprising given that insula-
of these or similar species using paired visible and TIR tion is generally a key property of bird nests (Mainwaring
cameras aim to capture imagery at similar resolutions to et al., 2014). Presumably this difference would have been
ours, that is ≤1 cm and ~5–10 cm for the visible and TIR less apparent earlier in the morning or under cloudy or
imagery, respectively, although modestly coarser resolu- overcast conditions. Readers should refer to Burke et al.
tions may also be workable if surveying large areas that (2019) for further guidance on optimizing animal detec-
require higher-altitude flights. It is also critical to use tion with drone-borne TIR cameras.
appropriate shooting settings (see Appendix S1) to pro-
duce sharp and adequately exposed images when using
Areas for improvement
cameras with customizable settings.
Regarding the thermal detectability of subjects, we con- One challenge of using thermal-infrared imagery is its
ducted drone surveys during the first half of the day (be- low pixel resolution which, beyond making it generally
fore solar noon) when there is greater contrast between difficult to detect small subjects, provides limited details
the warm-bodied birds (and the nests they sit on) and in each image that can be used to match and align

222 ª 2020 The Authors. Remote Sensing in Ecology and Conservation published by John Wiley & Sons Ltd on behalf of Zoological Society of London.
20563485, 2021, 2, Downloaded from https://zslpublications.onlinelibrary.wiley.com/doi/10.1002/rse2.183 by Cochrane Hungary, Wiley Online Library on [30/08/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
A. E. McKellar et al. Drones to Count Marshbirds

Figure 6. Four empty Black-crowned Night


Heron nests (green dots) from (A) visible
imagery and (B) thermal-infrared imagery
obtained from a drone flying at an altitude of
60 m over an abandoned colony at Quill Lake
South.

overlapping images in the mosaicking process. This is the sophisticated on-board image georeferencing systems are
reason we experienced difficulties mosaicking TIR images, available for drones, but at significant additional cost.
particularly in areas dominated by homogeneous open In addition, the use of computer-automated image
water. This issue could potentially be mitigated by analysis procedures may improve the efficiency of nest
increasing the overlap between images, although this detection and counting. A variety of techniques have been
would reduce survey efficiency due to the greater number employed for automatically detecting bird subjects in aer-
of images required to cover the same area, as discussed ial imagery (Chabot & Francis, 2016), although our study
above. Another possibility might be to deploy buoys represents a particularly challenging case because of the
throughout the survey area that could aid in image cryptic nature of some of the subjects, the need to distin-
matching, provided the buoys are discernable in the TIR guish among multiple co-occurring species in some colo-
imagery. nies as well as between active and inactive nests, and the
The use of these or other types of ground control points complementary reliance on two different types of ima-
(with known, recorded-on-the-ground coordinates) would gery. Nevertheless, increasingly sophisticated analysis
likely also reduce or eliminate misalignment of the visible workflows have been developed for these sorts of cases
and TIR mosaics that required manual correction during (e.g., Andrew & Shephard, 2017; Chabot et al., 2018;
post-processing. Then again, the additional time required Chrétien et al., 2015; Francis et al., 2020), and automated
to do this in the field would need to be weighed against analysis of remote sensing imagery in general appears to
the time required for correction in post-processing, not to be on the brink of being revolutionized by emerging
mention reintroduce some of the disturbance to colonies ‘deep learning’ algorithms, which have already been
associated with ground surveys. Alternatively, more employed for bird detection (Hong et al., 2019).

ª 2020 The Authors. Remote Sensing in Ecology and Conservation published by John Wiley & Sons Ltd on behalf of Zoological Society of London. 223
20563485, 2021, 2, Downloaded from https://zslpublications.onlinelibrary.wiley.com/doi/10.1002/rse2.183 by Cochrane Hungary, Wiley Online Library on [30/08/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
Drones to Count Marshbirds A. E. McKellar et al.

Table 1. Comparison of nest counts derived from dual visible-thermal imagery obtained during drone flights and from traditional ground surveys
at eight marshbird colonies on five lakes in Saskatchewan, Canada, in June and July 2019.

Flight % diff
Geographic altitude Date – Date – Count – Count – (drone –
Site coordinates (m) drone ground Species drone ground ground)

Foam Lake South 51.712, −103.582 60 17 June NA Franklin’s Gull 2337 NA NA


Quill Lake South1 51.9, −104.225 60 17 June 11-12 Black-crowned 1524 1582 −3.73
July Night Heron
Foam Lake North2 51.726, −103.589 120 18 June 23-24 Franklin’s Gull 1406 1466 −4.18
June
Good Spirit Lake 51.551, −102.629 45 18 June 17 June Black Tern 4 5 −22.22
Jackfish Lake East 53.131, −108.427 45 18 June 21 June Black Tern 6 7 −15.38
Jackfish 53.124, −108.436 45 19 June 23 & 26 Western Grebe 240 229 4.69
Lake West June Franklin’s Gull 154 152 1.31
Black Tern 4 3 28.57
Forster’s Tern 74 75 −1.34
Porter Lake 52.179, −106.291 60 19 June 25 June Black-crowned Night Heron 126 126 0
Quill Lake North 51.949, −104.242 60 21 June 21 June Western Grebe 109 113 −3.6

Percent difference between counts at each colony is shown, where negative differences indicate a lower count from the drone imagery. Note that
a ground survey was not conducted at Foam Lake South.
1
Colony was abandoned prior to drone and ground surveys, and all nests counted were empty.
2
Only a subset of the colony was surveyed during the ground survey (see Materials and Methods), and a count from the drone imagery of the
equivalent area is shown; nest count of the entire colony based on imagery was 9598.

imagery to detect cryptic nests and associated visible ima-


Implications and future work
gery to quantify microhabitat characteristics of nesting
Our results demonstrate promise for the use of drone- sites, which may prove more accurate and less disruptive
borne dual visible-thermal cameras for monitoring multi- than traditional on-the-ground nest vegetation surveys.
ple colonially breeding marshbird species in structurally
complex habitats. This type of survey could potentially be
Acknowledgements
applied to many other marsh-nesting species, given our
incidental and positive results for American Coots and The authors thank Luke Nych and Kayla Caruso for help
observations of Eared Grebes as well as several duck species in the field, and Pierre Audette for help piloting the
in the imagery. Indeed, Eared Grebes are difficult to accu- drone. Additionally, the authors thank the GIS support
rately survey during the breeding season using traditional and advice provided by Mark Bidwell, Zhong Li, and
methods due to the small size of their nests and their ten- Mike Watmough. Jon Pasher and Taylor Harmer gener-
dency to cover them with vegetation when approached by ously loaned their TIR camera for use in this study. The
humans or potential predators (Hanus et al., 2002). More authors also thank Colin Melanie for permitting us to
broadly, the dual camera approach could potentially be conduct surveys on his property. This study was funded
beneficial in any situation where wildlife subjects are diffi- by Environment and Climate Change Canada.
cult to survey with visible or TIR imagery alone. While our
goal was to utilize both types of imagery simultaneously to
Permits
aid in the detection of colonial marshbird nests, future
studies may wish to compare counts obtained only from Work was conducted under Scientific Permit 19-MC/SK-
visible imagery to those obtained from paired visible and SC002 from Environment and Climate Change Canada.
TIR imagery, to determine the relative benefits of a dual-
camera approach across different habitat types and for dif-
Data Availability Statement
ferent species. Another recent application of drone-based
TIR imagery involved the detection of cryptic grassland Because of the extremely large total file size of the ima-
songbird nests (Scholten et al., 2019). Pairing TIR with gery collected and analyzed in this study, we have not
high-resolution visible imagery as we did may enhance and uploaded it to a public repository, but can make arrange-
expand this type of application to other aspects of moni- ments to share it (along with other derived geospatial
toring and research. For example, one may wish to use TIR data files) by other means upon request.

224 ª 2020 The Authors. Remote Sensing in Ecology and Conservation published by John Wiley & Sons Ltd on behalf of Zoological Society of London.
20563485, 2021, 2, Downloaded from https://zslpublications.onlinelibrary.wiley.com/doi/10.1002/rse2.183 by Cochrane Hungary, Wiley Online Library on [30/08/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
A. E. McKellar et al. Drones to Count Marshbirds

Chabot, D. & Francis, C.M. (2016) Computer-automated bird


References
detection and counts in high-resolution aerial images: a
Amat, J.A. & Green, A.J. (2010) Waterbirds as bioindicators of review. Journal of Field Ornithology, 87, 343–359.
environmental conditions. In: Hurford, C., Schneider, M. Chrétien, L.-P., Théau, J. & Ménard, P. (2015) Wildlife
and Cowx, I. (Eds.) Conservation monitoring in freshwater multispecies remote sensing using visible and thermal
habitats. Netherlands: Springer. infrared imagery acquired from an unmanned aerial vehicle
Andrew, M.E. & Shephard, J.M. (2017) Semi-automated (UAV). ISPRS - International Archives of the
detection of eagle nests: an application of very high- Photogrammetry, Remote Sensing and Spatial Information
resolution image data and advanced image analyses to Sciences, 4, 241–248.
wildlife surveys. Remote Sensing in Ecology and Conservation, Elliott-Smith, E., Bidwell, M., Holland, A.E. & Haig, S.M.
3, 66–80. (2015) Data from the 2011 International Piping Plover
Barnas, A., Chabot, D., Hodgson, A., Johnston, D.W., Bird, Census: U.S. Geological Survey Data Series, 922, 296 p.
D.M. & Ellis-Felege, S.N. (2020) A standardized protocol for https://doi.org/10.3133/ds922
reporting methods when using drones for wildlife research. Environment and Climate Change Canada [ECCC]. (2013)
Journal of Unmanned Vehicle Systems, 8, 89–98. Bird conservation strategy for bird conservation region 11 in
Barr, J.R., Green, M.C., DeMaso, J. & Hardy, T.B. (2020) the Prairie and Northern Region: Prairie Potholes. Gatineau,
Drone surveys do not increase colony-wide flight behavior Quebec: Environment and Climate Change Canada.
at waterbird nesting sites, but sensitivity varies among Francis, R.J., Lyons, M.B., Kingsford, R.T. & Brandis, K.J.
species. Scientific Reports, 10, 3781. (2020) Counting mixed breeding aggregations of animal
Barr, J.R., Green, M.C., DeMaso, S.J. & Hardy, T.B. (2018) species using drones: lessons from waterbirds on semi-
Detectability and visibility biases associated with using a automation. Remote Sensing, 12, 1185.
consumer-grade unmanned aircraft to survey nesting Hanus, S., Wollis, H. & Wilkinson, L. (2002) Western
colonial waterbirds. Journal of Field Ornithology, 89, (Aechmophorus occidentalis) and Eared (Podiceps nigricollis)
242–257. grebes of Central Alberta: inventory, survey techniques and
Blanco, D.E. & Carbonell, M. (Eds.) (2001) The Neotropical management concerns. Species at Risk Report No. 41,
Waterbird Census. The first 10 years: 1990–1999. Memphis, Alberta Sustainable Resource Development, Fish and
USA: Wetlands International, Buenos Aires, Argentina & Wildlife Division, Edmonton, Alberta.
Ducks Unlimited, Inc. Heath, S.R., Dunn, E.H. & Agro, D.J. (2009) Black Tern
Brisson-Curadeau, É., Bird, D., Burke, C., Fifield, D.A., Pace, P., (Chlidonias niger), version 2.0. The Birds of North America
Sherley, R.B. et al. (2017) Seabird species vary in behavioural Online, 147.
response to drone census. Scientific Reports, 7, 17884. Hodgson, J.C., Baylis, S.M., Mott, R., Herrod, A. & Clarke,
Burger, J. & Gochfeld, M. (2009) Franklin’s Gull (Leucophaeus R.H. (2016) Precision wildlife monitoring using unmanned
pipixcan), version 2.0. The Birds of North America Online, 116. aerial vehicles. Scientific Reports, 6, 22574.
Burke, C., Rashman, M., Wich, S., Symons, A., Theron, C. & Hodgson, J.C., Mott, R., Baylis, S.M., Pham, T.T.,
Longmore, S. (2019) Optimising observing strategies for Wotherspoon, S., Kilpatrick, A.D. et al. (2018) Drones
monitoring animals using drone-mounted thermal infrared count wildlife more accurately and precisely than humans.
cameras. International Journal of Remote Sensing, 40, 439–467. Methods in Ecology and Evolution, 9, 1160–1167.
Butler, M.J., Harris, G. & Strobel, B.N. (2013) Influence of Hong, S.-J., Han, Y., Kim, S.-Y., Lee, A.-Y. & Kim, G. (2019)
whooping crane population dynamics on its recovery and Application of deep-learning methods to bird detection
management. Biological Conservation, 162, 89–99. using unmanned aerial vehicle imagery. Sensors, 19, 1651.
Chabot, D. & Bird, D.M. (2012) Evaluation of an off-the-shelf Hothem, R.L., Brussee, B.E. & Davis, W.E. Jr. (2010) Black-
unmanned aircraft system for surveying flocks of geese. crowned Night-Heron (Nycticorax nycticorax), version 2.0.
Waterbirds, 35, 170–174. The Birds of North America Online, 74.
Chabot, D. & Bird, D.M. (2015) Wildlife research and Howe, M.A., Geissler, P.H. & Harrington, B.A. (1989)
management methods in the 21st century: where do Population trends of North American shorebirds based on
unmanned aircraft fit in? Journal of Unmanned Vehicle the International Shorebird Survey. Biological Conservation,
Systems, 3, 137–155. 49, 185–199.
Chabot, D., Craik, S.R. & Bird, D.M. (2015) Population census Israel, M. & Reinhard, A. (2017) Detecting nests of lapwing
of a large Common Tern colony with a small unmanned birds with the aid of a small unmanned aerial vehicle with
aircraft. PLoS One, 10, e0122588. thermal camera. Proceedings of the 2017 International
Chabot, D., Dillon, C. & Francis, C.M. (2018) An approach Conference on Unmanned Aircraft SystemsMiami, FL,
for using off-the-shelf object-based image analysis software USA13–16 June 2017.
to detect and count birds in large volumes of aerial imagery. Jones, S.L. (2008) Western colonial waterbird survey protocols.
Avian Conservation and Ecology, 13, 15. U.S. Department of Interior, Fish and Wildlife Service,

ª 2020 The Authors. Remote Sensing in Ecology and Conservation published by John Wiley & Sons Ltd on behalf of Zoological Society of London. 225
20563485, 2021, 2, Downloaded from https://zslpublications.onlinelibrary.wiley.com/doi/10.1002/rse2.183 by Cochrane Hungary, Wiley Online Library on [30/08/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
Drones to Count Marshbirds A. E. McKellar et al.

Nongame Migratory Birds Coordinator’s Office, Denver, Sardà-Palomera, F., Bota, G., Viñolo, C., Pallarés, O., Sazatornil,
Colorado. V., Brotons, L. et al. (2012) Fine-scale bird monitoring from
Kays, R., Sheppard, J., Mclean, K., Welch, C., Paunescu, C., light unmanned aircraft systems. Ibis, 154, 177–183.
Want, V. et al. (2019) Hot monkey, cold reality: surveying Sauer, J.R., Pardieck, K.L., Ziolowski, D.J., Smith, A.C., Hudson,
rainforest canopy mammals using drone-mounted thermal M.-A., Rodriquez, V. et al. (2017) The first 50 years of the
infrared sensors. International Journal of Remote Sensing, 40, North American Breeding Bird Survey. Condor, 119, 576–593.
407–419. Scholten, C.N., Kamphuis, A.J., Vredevoogd, K.J., Lee-
LaPorte, N., Storer, R.W. & Nuechterlein, G.L. (2013) Western Strydhorst, K.G., Atma, J.L., Shea, C.B. et al. (2019) Real-
Grebe (Aechmophorus occidentalis). The Birds of North time thermal imagery from an unmanned aerial vehicle can
America Online. locate ground nests of a grassland songbird at rates similar to
Lyons, M.B., Brandis, K.J., Murray, N.J., Wilshire, J.H., traditional methods. Biological Conservation, 233, 241–246.
McCann, J.A., Kingsford, R.T. et al. (2019) Monitoring large Seymour, A.C., Dale, J., Hammill, M., Halpin, P.N. & Johnston,
and complex wildlife aggregations with drones. Methods in D.W. (2017) Automated detection and enumeration of
Ecology and Evolution, 10, 1024–1035. marine wildlife using unmanned aircraft systems (UAS) and
Magness, D.R., Eskelin, T., Laker, M. & Renner, H.M. (2019) thermal imagery. Scientific Reports, 7, 45127.
Evaluation of small unmanned aerial systems as a census Steinkamp, M., Peterjohn, B., Byrd, V., Carter, H. & Lowe, R.
tool for Aleutian Tern Onychoprion aleuticus colonies. (2003) Breeding season survey techniques for seabirds and
Marine Ornithology, 47, 11–16. colonial waterbirds throughout North America. Available
Mainwaring, M.C., Hartley, I.R., Lambrechts, M.M. & from: http://www.waterbirdconservation.org/pubs/PSGMa
Deeming, D.C. (2014) The design and function of birds’ nual03.pdf [Accessed 1 March 2020]
nests. Ecology and Evolution, 4, 3909–3928. Weimerskirch, H., Prudor, A. & Schull, Q. (2018) Flights of
McKellar, A.E., Nych, L.A. & Bidwell, M.T. (2019) Distribution drones over sub-Antarctic seabirds show species- and status-
and abundance of marsh-nesting colonial waterbirds in specific behavioural and physiological responses. Polar
Prairie Canada. Wilson Journal of Ornithology, 131, 583–597. Biology, 41, 259–266.
McNicholl, M.K., Lowther, P.E. & Hall, J.A. (2001) Forster’s Weseloh, D.V.C. (2011) Inland colonial waterbird and
Tern (Sterna forsteri), version 2.0. The birds of North marshbird trends for Canada. Canadian Biodiversity:
America. Number 595. Ecosystem Status and Trends 2010 Technical Thematic
Morris, R.D. (2006) Techniques for monitoring colonial Report No. 18, Canadian Councils of Resource Ministers,
waterbirds in the boreal forest. Technical Report, Canadian Ottawa, ON. iv +33 p.
Wildlife Service, Environment Canada, Prairie and Northern Witczuk, J., Pagacz, S., Zmarz, A. & Cypel, M. (2018)
Region, Edmonton, Alberta Exploring the feasibility of unmanned aerial vehicles and
Pöysä, H., Kotilainen, J., Väänänen, V.-M. & Kunnasranta, thermal imaging for ungulate surveys in forests -
M. (2018) Estimating production in ducks: a preliminary results. International Journal of Remote Sensing,
comparison between ground surveys and unmanned aircraft 39, 5504–5521.
surveys. European Journal of Wildlife Research, 64, 74.
Prescott, D.R.C., Unruh, J., Morris-Yasinski, S. & Wells, M.
(2018) Distribution and abundance of the Western Grebe
Supporting Information
(Aechmophorus occidentalis) in Alberta: an update. Alberta Additional supporting information may be found online
Species at Risk Report No. 160, Alberta Environment and in the Supporting Information section at the end of the
Sustainable Resource Development, Fish and Wildlife Policy article.
Branch, Edmonton, Alberta.
R Core Team. (2020) R: A language and environment for Figure S1. Three Franklin’s Gull nests (blue dots) from
statistical computing. Vienna, Austria: R Foundation for (A) visible imagery and (B) thermal-infrared imagery
Statistical Computing https://www.R-project.org/ obtained from a drone flying at an altitude of 120 m at
Reintsma, K.M., McGowan, P.C., Callahan, C., Collier, T., Foam Lake North.
Gray, D., Sullivan, J.D. et al. (2018) Preliminary evaluation Figure S2. Four Eared Grebe nests (orange dots) from (A)
of behavioral response of nesting waterbirds to small visible imagery and (B) thermal imagery obtained from a
unmanned aircraft flight. Waterbirds, 41, 326–331. drone flying at a height of 45 m at Jackfish Lake West. A
Rosenberg, K.V., Blancher, P.J., Stanton, J.C. & Panjabi, A.O. Franklin’s Gull nest can also be seen in the center.
(2017) Use of North American Breeding Bird Survey data in Table S1. Disturbance observations from drone flights
avian conservation assessments. Condor, 119, 594–606. over eight marshbird colonies in June 2019 in Saskatche-
Rosenberg, K.V., Dokter, A.M., Blancher, P.J., Sauer, J.R., wan.
Smith, A.C., Smith, P.A. et al. (2019) Decline of the North Appendix S1. Detailed drone-related materials and meth-
American avifauna. Science, 366, 120–124. ods.

226 ª 2020 The Authors. Remote Sensing in Ecology and Conservation published by John Wiley & Sons Ltd on behalf of Zoological Society of London.

You might also like