In situ evaluation of an automated aerial bait delivery system for landscape-scale control of invasive brown treesnakes on Guam

Abstract

After decades of biodiversity loss and economic burden caused by the brown treesnake invasion on the Pacific island of Guam, relief hovers on the horizon. Previous work by USDA Wildlife Services (WS) and its National Wildlife Research Center (NWRC) demonstrated that brown treesnake numbers in forested habitats can be dramatically
suppressed by aerial delivery of dead newborn mouse (DNM) baits treated with 80 mg of acetaminophen. However, manual bait preparation and application is impractical for landscape-scale treatment. WS, NWRC, and the US Department
of the Interior have collaborated with Applied Design Corporation to engineer an automated bait manufacturing and delivery system. The core technology is an aerially delivered biodegradable “bait cartridge” designed to tangle in the tree canopy, making the acetaminophen bait available to treesnakes and out of reach of terrestrial non-target organisms. When mounted on a rotary- or fixed-wing airframe, the automated dispensing module (ADM) unit can broadcast 3,600 bait cartridges at a rate of four per second and can treat 30 hectares of forest at a density of 120 acetaminophen baits per hectare within 15 minutes of firing time. We conducted the first in situ evaluation of the ADM in July 2016. Initial acetaminophen bait deployment rates (proper opening of the bait cartridge for canopy entanglement) were low, and mechanism jams were frequent due to internal friction and wind forces; on-site remedial engineering improved these performance measures. Bait cartridge placement and spacing were accurate (average 8.9 m along 9 m swaths) under various flight heights and speeds. Canopy entanglement of properly-deployed acetaminophen baits was high (66.6%). Although only a small proportion (5.9%) of radio transmitter-equipped acetaminophen baits were confirmed to have been taken by brown treesnakes, the baiting density was high enough to make it likely that a significant proportion of brown treesnakes in the area had taken acetaminophen baits. With subsequent improvements in system reliability, the automated bait cartridge manufacturing and delivery system is poised to transition from research and development to operational field implementation. Applications include reduction of brown treesnake numbers around transportation infrastructure and within core habitats for the reintroduction of native birds extirpated by this troublesome invasive predator.

Full text

348
INTRODUCTION
The brown treesnake (Boiga irregularis) is a nocturnal,
arboreal predator that was probably introduced on the
island of Guam after World War II as a passive stowaway
in cargo from the Admiralty Islands north of New Guinea
(Rodda & Savidge, 2007; Richmond, et al., 2014). Lacking
natural predators on Guam, the population of brown
treesnakes irrupted, reaching as many as 50–100 brown
treesnakes per hectare in some areas (Rodda, et al., 1999).
Brown treesnakes colonised the entire island of Guam
(54,930 ha) in about 20–30 years (Savidge, 1987). The
brown treesnake has been – and continues to be – a threat
to the economy and ecology of Guam, and is currently
the subject of a cooperative programme to control brown
treesnake populations on the island and prevent its spread
throughout the Pacic Basin and other vulnerable locations
(Clark, et al., 2018). Owing to the signicant ecological
and economic damages caused by the brown treesnake on
Guam, the potential for the brown treesnake to be spread to
other Pacic Islands, including Hawai`i, is of great concern
(Shwi, et al., 2010).
Landscape-scale suppression of brown treesnakes is
desirable in habitats adjacent to transportation network
infrastructure (e.g., cargo terminals), to reduce the risk
of accidental transport to other vulnerable ecosystems,
and within key habitats for the recovery of Guam’s native
wildlife. Because of the great amount of inaccessible and
topographically challenging forest habitat on Guam, aerial
delivery of brown treesnake suppression tools is key to the
management of this species on a landscape scale. Dead
newborn mice (DNM) dosed with 80 mg of acetaminophen
have proven to be safe and eective baits for lethal control
of brown treesnakes (Savarie, et al., 2001; Johnston, et al.,
2002; Clark, et al., 2012) and are registered with the US
Environmental Protection Agency (EPA) as an approved
pesticide (Registration No. 56228-24, Revised 06/2018).
To be eectively delivered to the forest canopy where they
are available to foraging brown treesnakes and inaccessible
by terrestrial non-targets, the baits must be coupled with a
‘otation device’ intended to entangle in foliage (Savarie
& Tope, 2004).
Through a previous project, the US Department of
Agriculture (USDA) Animal and Plant Health Inspection
Service (APHIS) Wildlife Services National Wildlife
Research Center (NWRC) has demonstrated that brown
treesnake abundance in Guam’s forests can be suppressed
via the aerial application of DNM baits adhered to paper
streamers (Dorr, et al., 2016). During this prior study, baits
were hand-prepared and hand-broadcast from a helicopter.
While this method of treatment proved eective on a
small scale (two 55 ha plots), manual bait preparation
and application is economically impractical for larger
landscape-scale treatments. In scaling up to meet the
challenge of landscape-scale control of brown treesnakes,
one of the principal logistical concerns is the obvious need
to automate both bait production and the aerial dispensing of
baits. In response to this need, NWRC, primarily funded by
the US Department of the Interior Oce of Insular Aairs,
has partnered with a private, small business engineering
company (Applied Design Corporation, Boulder, Colorado)
to develop a brown treesnake suppression system that oers
the capability to achieve precise distribution of thousands
of baits in a matter of minutes, through a fully-integrated
In situ evaluation of an automated aerial bait delivery system for
landscape-scale control of invasive brown treesnakes on Guam
S.R. Siers1, W.C. Pitt1,4, J.D. Eisemann1, L. Clark1, A.B. Shiels1, C.S. Clark2, R.J. Gosnell2 and M.C. Messaros3
1USDA APHIS Wildlife Services National Wildlife Research Center, LaPorte Ave, Fort Collins CO 80521.
<shane.r.siers@aphis.usda.gov>. 2USDA APHIS Wildlife Services, Western Region, Centre Ave, Fort Collins CO
80521. 3Applied Design Corporation, Western Ave, Boulder CO 80301. 4Current aliation: Smithsonian Conservation
Biology Institute, Remount Road, Front Royal, VA 22630.
Abstract After decades of biodiversity loss and economic burden caused by the brown treesnake invasion on the
Pacic island of Guam, relief hovers on the horizon. Previous work by USDA Wildlife Services (WS) and its National
Wildlife Research Center (NWRC) demonstrated that brown treesnake numbers in forested habitats can be dramatically
suppressed by aerial delivery of dead newborn mouse (DNM) baits treated with 80 mg of acetaminophen. However,
manual bait preparation and application is impractical for landscape-scale treatment. WS, NWRC, and the US Department
of the Interior have collaborated with Applied Design Corporation to engineer an automated bait manufacturing and
delivery system. The core technology is an aerially delivered biodegradable “bait cartridge” designed to tangle in the
tree canopy, making the acetaminophen bait available to treesnakes and out of reach of terrestrial non-target organisms.
When mounted on a rotary- or xed-wing airframe, the automated dispensing module (ADM) unit can broadcast 3,600
bait cartridges at a rate of four per second and can treat 30 hectares of forest at a density of 120 acetaminophen baits
per hectare within 15 minutes of ring time. We conducted the rst in situ evaluation of the ADM in July 2016. Initial
acetaminophen bait deployment rates (proper opening of the bait cartridge for canopy entanglement) were low, and
mechanism jams were frequent due to internal friction and wind forces; on-site remedial engineering improved these
performance measures. Bait cartridge placement and spacing were accurate (average 8.9 m along 9 m swaths) under
various ight heights and speeds. Canopy entanglement of properly-deployed acetaminophen baits was high (66.6%).
Although only a small proportion (5.9%) of radio transmitter-equipped acetaminophen baits were conrmed to have been
taken by brown treesnakes, the baiting density was high enough to make it likely that a signicant proportion of brown
treesnakes in the area had taken acetaminophen baits. With subsequent improvements in system reliability, the automated
bait cartridge manufacturing and delivery system is poised to transition from research and development to operational
eld implementation. Applications include reduction of brown treesnake numbers around transportation infrastructure
and within core habitats for the reintroduction of native birds extirpated by this troublesome invasive predator.
Keywords: invasive species suppression, invasive vertebrate predator, public-private partnership, scaling up, technical
innovation, toxic baits
S.R. Siers, W.C. Pitt, J.D. Eisemann, L. Clark, A.B. Shiels, C.S. Clark, R.J. Gosnell and M.C. Messaros
Siers, S.R.; W.C. Pitt, J.D. Eisemann, L. Clark, A.B. Shiels, C.S. Clark, R.J. Gosnell and M.C. Messaros. In situ evaluation of
an automated aerial bait delivery system for landscape-scale control of invasive brown treesnakes on Guam
In: C.R. Veitch, M.N. Clout, A.R. Martin, J.C. Russell and C.J. West (eds.) (2019). Island invasives: scaling
up to meet the challenge, pp. 348–355. Occasional Paper SSC no. 62. Gland, Switzerland: IUCN.

349
solution that encompasses bait cartridge production, an
aerial bait cartridge dispensing system, and supporting
infrastructure and logistics for practical manufacturing,
storing, and ight-line handling of bait cartridges.
Automated Bait Manufacturing System (ABMS)
Many of the functional details of the three-stage
ABMS are currently considered proprietary information
pending application for US and foreign patent protection.
The descriptions provided below will suce as a basic
functional explanation.
The rst of three bait cartridge manufacturing stations
is the Gluer/Placer Station (Station 1) where the DNM
are distributed on moulded pulp paper trays and an 80 mg
acetaminophen tablet is adhered to the DNM via a hot-
melt adhesive. At the nal stage of Station 1, the individual
capsules containing the acetaminophen tablet and DNM
are cut from the paper trays and fed into a transport cassette
for transfer to the Assembly/Winder Station (Station 2).
Hereafter, a DNM with an adhered acetaminophen tablet
will be referred to as an “acetaminophen bait”.
At Station 2, the capsule is folded and held closed
by pinching at the paper hinge between the two capsule
halves. This pinched paper hinge, hereafter referred to as
the ‘tang,’ is inserted into a slotted pressed pulp paper end
cap. One end of a biodegradable plastic ribbon is adhered
to the endcap and the entire assembly is rotated until the
ribbon is wound around the length of the capsule in a
‘barber pole’ fashion. The terminal end of the ribbon is
then adhered to the paper capsule. An exterior cardboard
tube is placed over the wound assembly, with the end cap
tightly pressed into the tube; this entire resulting assembly,
comprised of the acetaminophen bait, capsule, streamer,
and end cap, enclosed within the external tube, is referred
to as a “bait cartridge”. The entire bait cartridge (Fig. 1) is
biodegradable.
The nal manufacturing station is Packaging (Station
3). Completed bait cartridges are automatically fed to the
packaging station, where they are gathered and placed into
a corrugated plastic case (900 bait cartridges per case).
Filled cases are shrink wrapped, placed on a shipping pallet,
and frozen. A complete pallet of 40 cases holds 36,000
bait cartridges, enough to treat 300 ha at the current EPA-
approved maximum application rate of 120 acetaminophen
baits/ha.
Automated Dispensing Module (ADM)
The Automated Dispensing Module (ADM; Fig. 2) is
comprised of three main components: 1) four magazines;
2) an electro-mechanical ring unit on a tilt-plate; and 3) a
frame, which holds the power supply battery, the computer
control module, and integrates the other components into
a single functional ADM. The frame is mounted within the
hold of the aircraft.
Each magazine is comprised of a body with two halves
hinged at the back, allowing the payload area to be fully
exposed, and a faceplate. The opened magazine receives
the contents of one case (900 bait cartridges). Upon
loading, the bait cartridges receive a nal inspection for
manufacturing imperfections or shipping damage which
may adversely aect smooth feeding through the magazine
and into the ring unit (Fig. 3).
Fig. 1 When deployed, the bait capsule and outer bait
cartridge tube are joined by a length of unfurled ribbon
intended to entangle in the forest canopy when applied
aerially.
Fig. 3 Bait cartridges are inspected for manufacturing
imperfections or shipping damage that might impede
smooth feeding and ejection.
Fig. 2 The ADM is comprised of four ring units and four
900-cartridge magazines along with an onboard battery
and control electronics (not visible).
Siers, et al.: Automated bait system for treesnakes on Guam

350
A magazine can be loaded with a case of bait cartridges
and prepared for ight in two to three minutes. Once four
magazines are prepared, they are loaded into the aircraft-
mounted ADM frame (Figs. 4 and 5). The bait cartridge
exit door on each magazine is then opened, allowing
bait cartridges to ow into the ring unit feed chute. A
full payload of 3,600 bait cartridges is sucient to treat
30 ha of forest at the maximum application rate. At full
performance, this area can be treated at 120 acetaminophen
baits/ha within 15 minutes of ring time (Fig. 6). An
additional set of magazines allows for the next payload to
be prepared while the current payload is being applied.
A payload manager and the pilot are the only personnel
aboard the aircraft. As directed by the payload management
software, the computer control module engages the ring
units within the ADM to re bait cartridges at the proper
rate to match the aircraft’s current ground speed and
intended acetaminophen bait application rate. The payload
management software detects when a port is jammed or a
magazine is empty and increases the ring rate of the other
three ports to maintain the desired bait cartridge delivery
rate.
Aerial navigation is achieved by following a
preprogrammed mission plan in the payload management
software, which details the transects to be own. The pilot
is provided with an LCD display, a “virtual lightbar,” that
provides realtime feedback as to whether the aircraft is on
the prescribed ight path and what corrective movements
are needed to return to the path. The payload manager
manually toggles on bait cartridge ring when over the
treatment area, and toggles it o when the ight path is
complete. After an ‘ag turn’ (an aerial maneuver to quickly
reverse directions) the next ight path is own in the
opposite direction. This is repeated until the payload is
expended or the treatment area has been fully covered.
Objectives
This report describes the rst in situ evaluation of this
system through the experimental treatment of 110 hectares
of forest on Guam. The major objectives were to evaluate:
1) the ground support work ow and performance of the
automated dispensing module in-ight; 2) the precision of
spatial coverage of the treatment area; and 3) the proper
deployment of bait cartridges into the forest canopy and
the fate of acetaminophen baits once distributed into the
environment.
MATERIALS AND METHODS
Study site
The evaluation was conducted over 110 ha of
secondary forest on the Marbo Annex of Andersen Air
Force Base (typically referred to as “Andy South”) in
Yigo, Guam, at approximately 13.508°N, 144.873°E. This
site was selected because: 1) there is low risk to threatened
or endangered species; 2) the habitat is representative of
much of Guam’s forests; and 3) it is on a closed military
facility with restricted public access.
ADM performance
We assessed the performance of the ADM through
two trial applications of acetaminophen baits, simulating
operational applications for brown treesnake control. The
rst application was initially scheduled to be completed
on 19 July 2016, during which 13,200 acetaminophen
Fig. 5 Complete ADM with loaded magazines mounted in a
McDonnell-Douglass MD 500D helicopter.
Fig. 4 Magazines are loaded into the helicopter-mounted
ADM frame.
Fig. 6 Bait cartridges dispensed in ight.
Island invasives: scaling up to meet the challenge. Ch 2C Other taxa: Herpetofauna

351
baits would be applied over the 110 ha treatment area
(120/ha). A second application was scheduled to occur
three days later. For the purposes of this report, we
dene an “application” as a treatment of an area with
aerially-distributed acetaminophen baits within the usage
restrictions described in the EPA label.
A McDonnell-Douglas MD 500D (Fig. 5) and pilot were
contracted from Hansen Helicopters (Tamuning, Guam) to
perform aerial bait cartridge delivery. GoPro video cameras
(GoPro, Inc., San Mateo, California) were positioned
at various locations on the helicopter to document and
evaluate bait cartridge ejection and deployment success.
On the night prior to ight operations, bait cartridge
cases required for the next day’s application were removed
from the freezer to thaw and were stored overnight in an
air-conditioned workspace to minimise condensation. The
plastic wrapping on the cases were left intact to ensure that
all condensation would occur on the external surface of the
plastic wrap rather than on the paper-based bait cartridges
themselves.
Bait cartridge coverage
Bait cartridge spacing trials were conducted to
determine the accuracy and evenness of bait cartridge
distribution at varying ight heights and airspeeds. Three
lanes of approximately 200 m were delineated with orange
trac cones within an open grassy area at the treatment
site. The helicopter, traveling at 50 knots, distributed bait
cartridges along each ight line at heights of 25 m, 50 m,
and 100 m above ground level. A ground crew attempted
to locate all bait cartridges and measured their distance
from the ideal ight path and the distance to the next bait
cartridge along that path. A second round of transects was
own, this time at 60 knots, to determine the eect of
airspeed on accuracy and spacing.
The completeness and the evenness of the spatial
coverage of the treatment area was determined by recording
the GPS ight paths in the payload management software,
and generating coverage maps. Flight path segments were
highlighted where the ADM unit was ring.
Acetaminophen bait fate
Methods for monitoring of radio transmitter-equipped
baits were modied from procedures established by Dorr,
et al. (2016). During each treatment, a subset of baits was
prepared containing small 1.0 g VHF radio transmitters
(Holohil BD-2H with internal helical antennae, Holohil
Systems Ltd., Carp, Ontario, Canada) implanted in the
acetaminophen bait DNM. Transmitter-equipped bait
cartridges were placed directly in the ADM ring port
unit so that they would be deployed simultaneously at
the beginning of the ight path, to be followed by bait
cartridges without transmitters.
An acetaminophen bait is considered properly
“deployed” when the inner capsule assembly slides
out of the outer cardboard tube, unfurling the ribbon to
allow entanglement in the forest canopy. While some
acetaminophen baits may deploy on impact with treetops,
the system is designed for the acetaminophen bait to deploy
in the air immediately upon ejection of the bait cartridge
from the ADM.
Immediately after being aerially distributed, eld
technicians with handheld VHF receivers located the
transmitter-equipped baits and recorded: bait cartridge
location, position (in tree/vegetation or on ground), type of
vegetation the bait cartridge was suspended from, height
above ground, whether the bait cartridge was actually seen
or its location was estimated, whether the acetaminophen
bait was properly deployed and the DNM available for
take by a brown treesnake, whether the acetaminophen
tablet was still adhered to the mouse, and other notes about
the circumstances of the condition and location of the
acetaminophen bait and its availability for take by a brown
treesnake.
If a DNM became separated from the bait cartridge and
was on the ground but still had the acetaminophen tablet
attached, it was considered intact and available for take by
a brown treesnake. If the acetaminophen bait did not deploy
properly and the DNM was not available to be taken, the
bait cartridge and transmitter were recovered and that trial
was ended. After deployment-day data were collected, the
transmitter-equipped baits were left to determine the fate
of acetaminophen baits over the next 48–72 hours. On each
day following the application, each transmitter was re-
located and the following data were recorded: whether the
acetaminophen bait was still present and viable, whether
the acetaminophen tablet was still attached, whether the
acetaminophen bait was consumed by a brown treesnake
or a non-target, whether the brown treesnake or non-target
was alive or dead, whether the transmitter had moved to
a new location, and other notes about acetaminophen bait
location and condition.
If acetaminophen baits were unconsumed and still
viable, they were left for another night and located again
the next day. If acetaminophen baits had been consumed by
a brown treesnake or non-target that was still alive, it was
left undisturbed and relocated daily to establish survival
or time to death. While tracking transmitters, technicians
were alert for carcasses of any dead organisms, including
those that had ingested transmitter-equipped baits.
Global Positioning System (GPS) locations and notes on
the location and condition of carcasses were recorded.
Carcasses were collected and stored frozen for future
analytical chemistry to verify acetaminophen exposure.
RESULTS
ADM performance
The rst application of bait cartridges commenced
on schedule on 19 July 2016. Ground operations and
logistical support proceeded according to plan. However,
crew and video observations indicated poor ADM
performance in two primary categories: 1) bait cartridge
feed/ejection reliability and 2) percentage of bait cartridges
properly deploying in ight. These problems with system
performance resulted in frequent ight stoppages to
resolve bait cartridge jams and address other engineering
challenges. As a result, additional ight days on 20, 22, 23,
25, and 26 July were required to achieve the rst complete
coverage of the treatment area.
Reliable bait cartridge ejection was hampered in three
primary manners: 1) mechanical jams in the ring unit; 2)
“starvation” of the ring unit feed ramp (bait cartridges not
arriving at the ring position from the magazine); and 3)
impediment of ejection by aerodynamic forces. These are
not distinct processes, with multiple possible interactions
among them. These issues were resolved with a variety of
on-the-y eld improvements, with the causes and eects
noted for future ADM design improvements.
Acetaminophen baits that do not deploy from the bait
cartridge constitute a waste of resources (because the
toxicant is inaccessible to snakes) and a fruitless toxic
input into the environment. While we did not expect 100%
deployment, observations by ground crew and video
camera evidence indicated that initial acetaminophen bait
deployment rates were unacceptably low at far less than
50%. Acetaminophen bait deployment issues generally fell
Siers, et al.: Automated bait system for treesnakes on Guam

352
into two categories: inadequate rotational energy imparted
by the ring unit to overcome external air resistance eects
and internal friction between the sliding components of the
bait cartridge.
Air resistance eects were largely mitigated by
employing adjustable baes near the ring unit ejection
ports to disrupt ejection-inhibiting air currents. Internal
friction eects were traced to excessive friction between the
bait cartridge capsule ‘tang’ and end cap. As manufactured,
the tang (folded paper hinge) of the interior clamshell is
seated in the slot of the end cap to prevent rotation of the
internal assembly during manufacturing and unwinding of
the ribbon during shipping and handling. However, it was
discovered that the tension of the ribbon wound around
the clamshell capsule caused the inner assembly to rotate
slightly and the tang to twist against the sides of the slot
in the end cap. This friction, along with the taut wind of
the ribbon, created a ‘locking’ force, holding the entire
assembly together and resisting the available centrifugal
force which would otherwise deploy the acetaminophen
bait properly. We determined that tearing o the paper tang
would relieve the friction against the end cap slot, greatly
increasing the deployment rate. For the second application,
all bait cartridges were prepared by manual removal of the
paper tang.
After system modications were made, the second
application was re-scheduled for 29 July 2016 (three days
after the completion of the rst application in accordance
with EPA label restrictions). During this application, bait
cartridge ejection and acetaminophen bait deployment
were far more reliable. Bait cartridge jams in ring ports
were less frequent and were promptly cleared. The only
signicant delay occurred when an ejector unit bearing
broke; a temporary bushing replacement was fabricated
and the ADM was returned to service within a few hours.
Aside from this stoppage, the entire second application
was completed within 2.5 hours.
Even after the above-mentioned modications, only
37.3% of acetaminophen baits (571 out of a sample of
1,528 bait cartridge ejections observed on video) deployed
immediately, as intended. Bait cartridges could only
reliably be observed for about a third of their trajectory
to the canopy, and some certainly deployed lower in
the airstream. Still more would have deployed upon
impact with the canopy or the ground. Nonetheless, we
determined that improvement is needed in the reliability
of aerial deployment of acetaminophen baits. Though there
is no way to be certain of the actual deployment rates, we
presume the realised acetaminophen bait deployment rate
to be <50% for the overall acetaminophen bait application
period.
Bait coverage
Bait cartridge placement and spacing was tested on 28
July 2016. Wind conditions during all ights were recorded
at 0 to 1 on the Beaufort scale (0 = < 1 km/h, calm, smoke
rises vertically; 1 = 1–5 km/h, light air, wind motion visible
in smoke). When air movement was detectable, it was
moving north to north-northwest. Flight direction was west
to east or east to west. Bait cartridge distributions over trial
ight paths are depicted in Fig. 7.
Placement along target ight paths and within 9-m
swaths was very accurate and consistent. The one exception
was the run at 100 m ight height at 50 knots airspeed; these
results are inconsistent with the other ve, and we consider
this to be an anomalous lapse in pilot ight accuracy.
Results do not appear to be inuenced by the dierence
between 50 and 60 knots airspeed. Likewise, accuracy of
placement along paths did not appear to be inuenced by
ight height. The most challenging combination of higher
ight speed (60 knots) and highest ight height (100 m)
resulted in an acceptable distribution pattern. Spacing
between bait cartridges along a given ight path was
highly variable, but the mean overall spacing of 8.9 m was
virtually identical to the target spacing of 9 m.
Due to frequent ight stoppages during the rst
application, the full site coverage was achieved piecemeal
over several days, with the entire area being treated by
the 6th ight day. While the appropriate number of bait
cartridges was deployed, the evenness of transect spacing
was of reduced importance compared to overcoming
the engineering challenges. The second application of
acetaminophen baits was relatively uninterrupted. Flight
paths were own as planned which, along with increased
pilot and payload manager experience, resulted in a much
more even treatment (Fig. 8).
Acetaminophen bait fate
On 26 July 2016, 28 transmitter-equipped baits were
broadcast over the treatment site. On 29 July 2016, an
additional 23 were broadcast, for a total of 51 transmitter-
equipped baits. The conditions of acetaminophen baits on
the day of deployment are summarised in Table 1. Of the
Fig. 8 Flight paths from the second bait application. Green
swaths (portion of the ight paths where bait ring was
actuated) are depicted at 9 m width, the optimal bait
cartridge spacing for the 120 acetaminophen baits/ha
application rate.
Fig. 7 Bait cartridge spacing and placement results.
The centre line for each ight path indicates the target
line, over which the pilot ew and bait cartridges were
dispensed. Green boxes around the centre lines indicate
4.5 m on each side of the centre line, for a 9 m swath
(the ideal ight path spacing for applications at 120
baits/ha). “Spacing” is the average distance from one
bait cartridge to the next one along the ight path (target
spacing was 9 m).
Island invasives: scaling up to meet the challenge. Ch 2C Other taxa: Herpetofauna

353
51 transmitter-equipped baits, 92.2% deployed from the
bait cartridges, with the acetaminophen baits available to
brown treesnakes.
Thirty-four bait cartridges (66.6%) tangled in the canopy
as intended (Fig. 9). Thirteen (25.5%) were on the ground,
but open and available to be taken by ground-foraging
brown treesnakes. Four (7.8%) were not deployed (closed)
on the ground, making the bait and toxicant unavailable
to the brown treesnake. During each application, one
transmitter-equipped cartridge was in the canopy but
could not be conrmed to have deployed; we consider it
unlikely that an unopened bait cartridge would be caught
in the canopy, so assumed that these acetaminophen baits
deployed.
Of the 47 opened bait cartridges, two from each
application had the acetaminophen tablet detached from
the DNM, making it an ineective acetaminophen bait.
In total, 41 of the 51 acetaminophen baits (80.4%) had
acetaminophen tablets attached to the DNM and were
available for take by a brown treesnake. This should be
viewed as the overall successful bait deployment rate for
this sample of baits.
Of the 51 transmitter-equipped baits, canopy height
and acetaminophen bait height data were available on 33
acetaminophen baits that successfully deployed (18 from
Application 1 and 15 from Application 2). The hanging
height of the acetaminophen bait with respect to the
canopy height is represented graphically in Fig. 10. In a
linear regression, canopy height and acetaminophen bait
height were signicantly correlated (p << 0.001, adjusted
R2 = 0.694; the four bait cartridges on the ground were not
included in the regression). These results show that the
majority of deployed acetaminophen baits were entangled
within a few metres of the top of the canopy.
Deployed and intact acetaminophen baits were re-
checked daily, with very few conrmed takes by brown
treesnakes or non-target organisms (Table 2). Of the 51
transmitter-equipped baits, three (5.9%) were conrmed
by visual sighting to have been taken by brown treesnakes,
or 7.3% of the 41 transmitter-equipped baits known to
be available and intact. The 95% binomial condence
interval (logit parameterisation) for the estimated take
rate of 5.9% is 1.9% to 16.7%; given the small number
Fig. 10 Hanging height of the bait cartridge (y-axis)
in relation to the height of the canopy at that location
(x-axis).
Fig. 9 Desired canopy entanglement and acetaminophen
bait exposure.
Bait
cartridge
status
Application
1 (n=28) Application
2 (n=23) TOTAL
(n=51)
Opened in
canopy* 19 (67.9%) 15 (65.2%) 34 (66.6%)
Opened on
ground 7 (25.0%) 6 (26.1%) 13 (25.5%)
Not
deployed 2 (7.1%) 2 (8.7%) 4 (7.8%)
Unknown 1 (3.6%) 1 (4.3%) 2 (3.9%)
Total
deployed* 26 (92.3%) 21 (91.3%) 47 (92.2%)
Total known
deployed
and intact** 23 (82.1%) 18 (78.3%) 41 (80.4%)
Table 1 Status of transmitter-equipped bait cartridges
following ejection from ADM. “Deployed” means the
inner capsule completely exited the outer tube and the
acetaminophen bait was available for take by a brown
tree snake. “Intact” means the acetaminophen tablet
was still attached to the bait mouse and available to be
taken by a brown treesnake.
*Assumes that “unknown” bait cartridges in canopy were
deployed; **Does not assume “unknown” baits were intact.
Species Application
1Application
2Total
Brown tree snake 1 2 3
Monitor lizard 1* 0 1
Marine toad 0 2* 2
Unknown 0 1 1
Table 2 Transmitter-equipped acetaminophen baits taken
by target (brown treesnake) or non-target species.
*Transmitter recovered in faeces
Siers, et al.: Automated bait system for treesnakes on Guam

354
of acetaminophen baits equipped with transmitters, the
actual rate of acetaminophen bait take by brown treesnakes
could vary widely. All three transmitters were regurgitated
prior to death, so no transmitters were recovered in brown
treesnake carcasses. All three transmitters taken by non-
targets were later found in faeces; it is unclear whether any
of these animals succumbed to acetaminophen toxicosis.
All vertebrate carcasses encountered during eld
activities were collected. This included three brown
treesnakes and one marine toad (Rhinella marina).
DISCUSSION
ADM performance
Future improvements to the ADM will focus on:
baing of the airstream around the ejector ports to
prevent interference with ejection; improved feeding of
bait cartridges from redesigned magazines and; increased
energetic impact imparted to the bait cartridge at the instant
of ring in order to improve ejection and deployment
reliability. Engineering modications to the ABMS
will further address the non-deployment issue through
tighter quality control on bait cartridge imperfections and
abatement of tang friction through a redesigned end cap.
Bait cartridge coverage
The accuracy of bait cartridge placement along ight
lines was encouraging. There was very little air movement
during these trials; under windier conditions, bait cartridges
distributed from greater heights will be more likely to drift
further from the intended ight path.
We attribute high variability in bait cartridge spacing
along the ight lines to variability in the times at which
acetaminophen baits deployed after being ejected from the
ADM. When the acetaminophen bait deploys, wind drag
increases greatly and the forward momentum is quickly
attenuated, causing the bait cartridge to drop straight
down. Acetaminophen baits that deploy later maintain
forward momentum longer and will move farther along the
ight path before landing. It is expected that bait cartridge
modications that improve acetaminophen bait deployment
will also result in less variability in time of deployment,
leading to more consistent spacing along ight paths.
Variability in spacing along the ight path does have
the potential to aect bait cartridge placement accuracy
at the edges of treatment areas where bait cartridge
application begins and ends, potentially leading to a
small number of bait cartridges landing outside of the
desired treatment area. To make up for the inconsistency
of bait cartridge density at these edges, it is advisable that
another application ight should occur along these edges,
perpendicular to the original ight paths, ensuring that the
edges get a full treatment in a more controlled fashion,
similar to coastal aerial rodenticide applications during
island rodent eradications.
Variability in bait cartridge placement along and
perpendicular to the ight path will add apparently random
“noise” to the locations, as opposed to placing bait cartridges
precisely on an idealised 9 x 9 m grid. This variability will
not aect the ability to get acetaminophen baits into the
movement areas of every brown treesnake. The greatest
risk of gaps in coverage might arise from strong changes in
wind direction, which might introduce strong biases in bait
cartridge drift patterns. This will likely factor in with other
considerations leading to recommendations not to apply
baits during high wind conditions.
Wind eects at bait cartridge ejection ports and direct
sunlight on the bait cartridge counter photogates resulted
in unreliable bait cartridge counts as tabulated by the ADM
onboard software. We ensured that EPA label application
rate restrictions were not exceeded by conrming that no
more than 14.66 cases (13,200 bait cartridges) were applied
throughout the treatment area during each application
period.
Acetaminophen bait fate
The proportion of transmitter-equipped baits taken by
brown treesnakes was low (5.9%); however, only a very
small portion of the bait cartridges distributed (0.19%)
were equipped with transmitters. If we assume that half of
the acetaminophen baits applied during both applications
properly deployed and were viable, then there were 13,200
acetaminophen baits available for take by brown treesnakes.
If 5.9% of those acetaminophen baits were taken by brown
treesnakes, we would expect approximately 779 brown
treesnakes to have taken an acetaminophen bait. If we
assume a density of 25 brown treesnakes per hectare in this
area (a conservative estimate based on the 25-50/ha range
reported by Rodda, et al. 1999), 2,750 brown treesnakes
would have been exposed to the treatment. If 779 brown
treesnakes took acetaminophen baits and were killed, this
would be a brown treesnake mortality of approximately
28% in what was eectively a single treatment (given
the low deployment rate). The three acetaminophen baits
visually conrmed to have been taken by brown treesnakes
were found on the ground, apparently regurgitated. In
previous NWRC lab ecacy trials of acetaminophen
baits with acetaminophen tablets internally-implanted in
the DNM (rather than glued to the exterior), 26% were
regurgitated, but 100% of the caged brown treesnakes that
regurgitated the acetaminophen bait died within 12 to 36
hours (Savarie, 2002). Based on that result, it is reasonable
to assume that the brown treesnakes that had taken and
regurgitated acetaminophen baits with transmitters in this
study also died.
With respect to deployment and entanglement rates,
caution should be taken in considering transmitter-
equipped cartridges to be representative of the standard
bait cartridges distributed during this evaluation. Machine-
assembled bait cartridges were manually unwound and
rewound by hand after the implantation of the radio
transmitter; this may explain why transmitter-equipped
cartridges deployed at a higher rate than those observed
on video. The added mass of the transmitter may also have
an eect on the forces exerted on various parts of the bait
cartridge and acetaminophen bait assembly. However,
it is also possible that unopened bait cartridges without
transmitters actually did deploy lower in the air column
(out of view of the video cameras) or upon impact with
the canopy.
The overall reduction of brown treesnake abundance in
the treatment area – as inferred from a foraging activity
index based on take rates of non-toxic DNM from bait
stations – is currently being monitored as a separate study
for future publication.
CONCLUSION
Upon ring from the ADM, bait cartridge ejection
and acetaminophen bait deployment reliability was
initially low. Performance was improved dramatically
with eld-improvised remedial measures. It is estimated
that <50% of acetaminophen baits deployed from the bait
cartridges, resulting in an under-treatment compared to the
target application rate of 120/ha. Canopy entanglement
of acetaminophen baits that properly deployed was
high. Aerial bait cartridge placement and spacing were
satisfactorily accurate. Reliability of bait cartridge ejection
Island invasives: scaling up to meet the challenge. Ch 2C Other taxa: Herpetofauna

355
and acetaminophen bait deployment will be a critical focus
of bait manufacturing and delivery system improvements,
increasing per-cartridge eectiveness. Future advancements
of this technology may include adaptation for payload
management by the pilot alone, incorporation of a longer-
lasting articial bait to replace the DNM, and increases in
ejector unit and magazine capacity for greater payloADM.
With this evaluation – and subsequent improvements in
system reliability – we consider the concept of automated
bait production and aerial delivery to be fundamentally
sound. For the rst time in the decades-long saga of
the brown treesnake invasion of Guam, the prospect of
landscape-scale suppression hovers on the horizon.
DISCLAIMER
The use of trade or corporation names within this report
is for the convenience of the user in identifying products.
Such use does not constitute an ocial endorsement
or approval of any product by the U.S. Department of
Agriculture.
ACKNOWLEDGEMENTS
We wish to thank the DOI Oce of Insular Aairs
for funding the development of the ABMS and ADM
technologies, and for funding this evaluation. Additional
funding for this evaluation was provided by US Navy, Joint
Region Marianas. The ABMS and ADM systems were
conceived and built by ADC engineers Jonathan Fragoso,
Shane Vogt, Grady Barfoot, and Michael Messaros. Bill
Coon engineered the ADM software and ew as the
navigator and payload manager. Hansen Helicopters
general manager (Rufus Crowe), pilot (Dan O’Brien), and
service shop sta were instrumental in the success of this
evaluation. Methodology for acetaminophen bait fate via
radio transmitters was established by Brian Dorr (Dorr, et
al., 2016). Logistics and eld support were provided by
Francine Chlarson, Derek Hendricks, Jerome Larimer,
Joe Rabon, Anthony Thompson, and Rachel Volsteadt.
Additional site support on ight days was provided by
Ray Quichocho, Rico Terrazas, and Marc Hall. Regulatory
and permitting assistance were provided by Shannon
Hebert, James Watkins, Earl Campbell, Diane Vice, Jim
McConnell, and the Guam Environmental Protection
Agency.
AUTHOR CONTRIBUTIONS
S. Siers was the principal investigator for the eld
evaluation, establishing the study design, coordinating
site access and environmental compliance, curating data,
executing analyses and data visualisation, and writing the
original draft of the manuscript. S. Siers, M. Messaros, C.
Clark, and R. Gosnell supervised eld evaluation activities.
W. Pitt and M. Messaros conceptualised the automated
bait cartridge manufacturing and delivery systems. M.
Messaros was the chief engineer and developer of the
intellectual property associated with the bait cartridge and
associated manufacturing and delivery systems. W. Pitt,
L. Clark, A. Shiels, J. Eisemann, and S. Siers contributed
to programme administration, funding acquisition, and
other matters associated with research and development
of the automated systems. J. Eisemann coordinated
pesticide registration and technology transfer matters, and
contributed to on-site coordination of evaluation activities.
All authors contributed to review and editing of the
manuscript.
REFERENCES
Clark, L., Savarie, P.J., Shivik, J.A., Breck, S.W. and Dorr, B.S. (2012).
‘Ecacy, eort, and cost comparisons of trapping and acetaminophen-
baiting for control of brown treesnakes on Guam’. Human–Wildlife
Interactions 6: 222–236.
Clark, L., Clark, C.S. and Siers, S.R. (2018). ‘Brown tree snakes:
methods and approaches for control.’ In: W.C. Pitt, J.C. Beasley and
G.W. Witmer (eds.) Ecology and Management of Terrestrial Vertebrate
Invasive Species in the United States, pp. 107–134. New York: Taylor
& Francis.
Dorr, B.S., Clark, C.S. and Savarie, P. (2016). Aerial application of
acetaminophen-treated baits for control of Brown treesnakes (RC-
200925; NWRC study number: QA-1828). Environmental Security
Technology Certication Program Final Report, Virginia, USA:
Department of Defense, Alexandria.
Johnston, J.J., Savarie, P.J., Primus, T.M., Eisemann, J.D., Hurley, J.C.
and Kohler, D.J. (2002). ‘Risk assessment of an acetaminophen baiting
program for chemical control of brown tree snakes on Guam: Evaluation
of baits, snake residues, and potential primary and secondary hazards’.
Environmental Science and Technology 36: 3827–3833.
Richmond, J.Q., Wood, D.A., Stanford, J.W. and Fisher, R.N. (2014).
‘Testing for multiple invasion routes and source populations for the
invasive brown treesnake (Boiga irregularis) on Guam: Implications
for pest management’. Biological Invasions 17: 337–349.
Rodda, G.H., McCoid, M.J., Fritts, T.H. and Campbell III, E.W. (1999).
‘Population trends and limiting factors in Boiga irregularis.’ In: G.H.
Rodda, Y. Sawai, D. Chiszar and H. Tanaka (eds.) Problem Snake
Management: the Habu and the Brown Treesnake, pp. 236–254. New
York: Cornell University Press.
Rodda, G.H. and Savidge, J.A. (2007). ‘Biology and impacts of Pacic
island invasive species: 2. Boiga irregularis, the brown tree snake
(Reptilia: Colubridae)’. Pacic Science 61: 307–324.
Savarie, P.J., Shivik, J.A., White, G.C., Hurley, J.C. and Clark, L. (2001).
‘Use of acetaminophen for large-scale control of brown tree snakes’.
Journal of Wildlife Management 65: 356–365.
Savarie, P.J. (2002). Acute oral toxicity of acetaminophen tablets to brown
treesnakes, Unpublished report – QA-636. Fort Collins, Colorado:
National Wildlife Research Center.
Savarie, P.J. and Tope, K.L. (2004). ‘Potential otation devices for
aerial delivery of baits to brown treesnakes’. In: R.M. Timm and W.P.
Gorenzel (eds.) Proceedings of the 21st Vertebrate Pest Conference, pp.
27–30. Davis, California: University of California.
Savidge, J.A. (1987). ‘Extinction of an island forest avifauna by an
introduced snake’. Ecology 68: 660–668.
Shwi, S.A., Gebhardt, K., Kirkpatrick, K.N. and Shwi, S.S. (2010)
‘Potential economic damages from introduction of brown tree snakes,
Boiga irregularis (Reptilia: Colubridae), to the islands of Hawai’i’.
Pacic Science, 64: 1–10.
Siers, et al.: Automated bait system for treesnakes on Guam