Using tracking technology to locate endangered ʻ ua ʻ u or Hawaiian Petrel ( Pterodroma sandwichensis ) burrows

,


INTRODUCTION
Searching for breeding colonies of nocturnal burrow-nesting seabird species can be particularly challenging, especially if the species is very rare, widely dispersed, and/or breeding in remote areas with dense vegetation and challenging topography. These factors are compounded for nocturnal burrow nesting species. However, locating these colonies is a critical first step to protecting them, as threats (such as introduced predators) within colonies often need intensive and directed management (Croxall et al. 2012, Gummer et al. 2015, Rayner et al. 2015, Raine et al. 2020a.
In this paper, we outline the use of telemetry data to locate a breeding colony of the endangered ʻuaʻu, or Hawaiian Petrel Pterodroma sandwichensis.
The Hawaiian Petrel is endemic to the Hawaiian Islands. Although Hawaiian Petrels bred historically on all main Hawaiian Islands (Olson and James 1982), their populations have contracted in recent decades with the majority of breeding birds now found on the islands of Kauaʻi, Maui, and Lānaʻi (Pyle and Pyle 2017). The reason for these declines include collisions with powerlines (Cooper and Day 1998, Podolsky et al. 1998, Raine et al. 2017a, Travers et al. 2021, introduced predators (Simons 1985, Hu et al. 2001, 2020a, and habitat modification within breeding colonies caused by invasive plants and ungulates (Duffy 2010, VanZandt et al. 2014. A small number of fledglings are also attracted annually to artificial lights, although this is not as important an issue as it is for another endangered seabird on Kauaʻi-the ʻaʻo or Newell's Shearwater Puffinus newelli (Reed et al. 1985, Telfer et al. 1987, Raine et al. 2017a). This combination of factors has led to the Hawaiian Petrel being globally listed under the IUCN Red Data List as "Vulnerable" (BirdLife International 2022) and "Endangered" under the U.S. Endangered Species Act.
On Hawaiʻi, prior to 2011, Hawaiian Petrels were only known from leeward Mauna Loa in the Hawaiʻi Volcanoes National Park (Swift and Burt-Toland 2009). During initial fence construction in 2009 on a new conservation unit within the Puʻu O ʻUmi Natural Area Reserve (NAR), fencing staff reported hearing "odd sounds" during the night. This led to surveys being conducted on the north coast of Kohala mountain in 2011 and Hawaiian Petrels were confirmed through auditory and visual observations (Deringer 2012). Subsequent work using acoustic monitoring devices identified vocal activity of both Hawaiian Petrel and Newell's Shearwater in the area (Dunleavy and McKown 2018), but prior to this study active burrows had not been located, making it hard to focus limited resources on critical management actions.
To build upon previous survey work, we focused on physically locating active Hawaiian Petrel burrows within Puʻu O ʻUmi NAR ( Fig. 1) by capturing transiting (commuting) adults during the breeding period and attaching transmitters to them to track them back to their burrows. This technique has been used successfully in New Zealand with multiple nocturnal burrow-nesting seabirds including the Chatham Island Taiko Pterodroma magenta (Imber et al. 1994), the Chatham Petrel Pterodroma axillaris (Gummer et al. 2015), and the diminutive New Zealand Storm-Petrel Fregetta maoriana (Rayner et al. 2015). This method has also been used to locate the nests of Marbled Murrelets Brachyramphus marmoratus breeding in mature old growth forests in Oregon (Northrup et al. 2018) and to locate breeding grounds of the critically endangered Beck's Petrel Pseudobulweria becki (Rayner et al. 2020).

Study site
All fieldwork was carried out in the Puʻu O ʻUmi NAR on the island of Hawaii. This NAR was established in 1987 and consists of 4104 hectares, incorporating the wet summit lands of Kohala Mountain, montane bogs, ʻōhiʻa Metrosideros polymorpha dominated forests, shrublands, and grasslands. The Reserve extends from the summit of Kohala at 1890 m above sea level (asl) to the towering Kohala sea cliffs and a coastal dry grassland. Tagging was undertaken at an elevation of 914 m asl within an area of Puʻu O ʻUmi NAR known as Slant Camp (Fig. 1), which was where concentrated transiting activity (birds moving from the sea to inland breeding sites) by Hawaiian Petrels had previously been observed during auditory surveys carried out with night vision (Wang, personal observation). This area is bordered by the vertical and largely inaccessible cliffs of Waipiʻo and Waimanu valleys, which drop thousands of feet to the valley floors. Numerous streams run through the area, draining out of the flat bogs at the apex of the site where the camp is situated into steep sided drainages covered in dense uluhe ferns Dicranopteris linearis and ʻōhiʻa trees. The area receives high levels of rainfall and is often blanketed in heavy mist.

Tracking
Tagging was undertaken in the months of July and August (15-20 July 2018 and 26-30 August 2019) to ensure that petrels had hatched their eggs and were now feeding chicks, and to increase the chances that captured birds were breeders and not prospectors (Simons 1985, Judge 2011. This was an important facet of the work, because prospecting birds and non-breeders are a component of the population that would be visiting the colony to initiate pairing but would not be associated with an active burrow. Birds transiting over to undiscovered breeding sites were attracted using lights and speakers playing Hawaiian Petrel calls. In the first year, BIGSUN 2018 Super Bright LED Searchlights were used. In the second year, more powerful lights (Grandlumen 240W High Bay LED UFO shop Light) were used to increase catching efficiency. For some birds, once they started circling the lights, powerful flashlight beams were used to try to draw them toward the ground. Toward the end of the study, we stopped using flashlights as sufficient birds were landing next to the lights without this additional lighting source. Mist nets (12 m x 6 m) were also set up adjacent to the speakers and flood lights in the first year; however, these were not ultimately useful for catching petrels circling the area (no bird was ever caught during that year with a net) and were also discontinued in 2019.
On two dates over the study period, we tagged six Hawaiian Petrels, with two birds tagged in 2018 and four in 2019. Birds were tagged with data loggers (e-obs Bird Solar 10g -http://www.e-obs. de/). We chose the e-obs tags because they were the lightest depth-reinforced GPS tags that would give us the highest resolution data over multiple foraging trips for the study (tags were set to record locational data every minute if the tags had full power and four minutes if the battery dropped below a certain threshold). High resolution data was a critical element to help pinpoint active Hawaiian Petrel burrows. Although the manufacturers state that these tags provide GPS level accuracy (95% within a 10 m radius of the point), we had also previously tested the altitudinal accuracy of these tags when using them on breeding birds from known burrows of the same species on Kauaʻi. During this previous work, after data was collected from deployed tags we assessed the altitude of logger points where (i) it was confirmed that the bird was on the ground (based on speed and visual inspection of each point in relation to other consecutive points), and (ii) points were located within 10 m of the bird's burrow. Altitudinal readings were then compared to altitude measurements taken at burrows via a handheld GPS (using Garmin Rino 650s). The average difference between point altitude averages and elevations recorded at the burrow was 10.8 m ± 3.1 SE (n = 9, min = 0.1 m, max = 23 m). Based on the results of this previous analysis, it appears that the altitude reading of the tags is relatively accurate with only a small potential error and could indeed be used to assess whether birds were airborne or grounded.
A second important consideration for using these tags was that data from tags could be remotely downloaded to a base station (purchased directly from e-obs), which meant that the birds did not have to be recaptured, something that would be almost impossible considering that burrow locations were not known at the time of tagging, and the chances of drawing the same bird down using the light wells would be extremely low. The modified tags were the lightest units of this nature available and were 1.9% to 3.8% of petrel body mass (depending on the bird tagged), below the maximum recommended mass for devices attached to Procellariid seabirds (Phillips et al. 2003).
We weighed all study birds (± 1.0 g), checked them for signs of brood patches (classified as "full," "partial," or "none," although it should be noted that in 2019 birds were caught in late August when brood patches could have easily feathered over), and banded them with a stainless-steel band (size 3A, USGS Bird Banding Laboratory). We acknowledge that the use of brood patches for identifying breeding birds can be potentially deceptive as in some species (particularly storm-petrels) non-breeders can develop brood patches (e.g., Wilson's Storm-Petrel Oceanites oceanicus; Beck and Brown 1972). However, it is still considered to be a useful data point for providing an indication of breeding activity (e.g., in New Zealand Storm-Petrel; Rayner et al. 2015) and one we felt important to consider. After banding, birds were held by an experienced seabird handler (AW) and their heads covered by a light-weight cloth to shield them from light and keep them calm during tag attachment. The tag profile (~2.5 cm²) represented approximately 3% of the frontal area of a Hawaiian Petrel. We acknowledge that the increase in cross-sectional area could, to some degree, affect the hydrodynamics of Hawaiian Petrel, although because this species is not known to undertake deep dives during foraging Flora 2010, Morra et al. 2019), this was of less concern than weight burden and balance. In anticipation of potentially long-distance flights, we preferred to attach the tags consistent with the bird's center of mass to minimize interference with flight, balance, and behavior (Healy et al. 2004, Vandenabeele et al. 2014. Tags were programmed to transmit continuously every 60 s, with no off cycle. All tags were attached by AFR. In 2018, two different tag attachments were trialed. For one bird the data logger was attached to its back. We used a modified suture-tape-glue attachment technique (MacLeod et al. 2008, Jodice et al. 2015, Raine et al. 2020b) to attach the tag. Specifically, several feathers on the central, dorsal surface between the scapulae were lifted and one strip (0.5 × 2.0 cm) of waterproof tape (Tesa® 4651, Beiersdorf, Germany) was inserted adhesive-side-up and wrapped over on itself to secure several feathers. The tape served to mark the location where the center of the tag would sit. We used four, sterile surgical sutures (2-0 Prolene™ monofilament, nonabsorbable sutures, Ethicon, Somerville, NJ) to attach a 3Dprinted plastic platform to the skin. For each suture, the skin below the platform's custom suture holes was pinched using the thumb and fore finger, a sterile 21 gauge × 3.8 cm hypodermic needle was inserted through the pinched skin, and the suture then threaded through the needle. When the needle was removed, the suture was retained under a 17 mm wide section of skin (equivalent to the width of the base of the tag). The sutures were then threaded back through the holes along the edge of the platform and secured snug to the skin and feathers with four surgical square knots. Care was taken to ensure that each suture was snug to minimize risk to the bird for entanglement, which could be a concern for birds returning to burrows. The tag was then attached to the plate using three small cable ties (Fig. 2). For the second bird in 2018 we attached the tag to the tail. The 3D printed base was taped using Tessa tape to two central tail feathers and the data logger cable-tied to the mount. This species has a relatively long tail and it was thought that a tail mount might be appropriate and stable for the petrels and could also increase tag attachment duration. Because of the small amount of data collected by this method in 2018 (and with similar results using larger samples sizes on Kauaʻi and Lānaʻi; Raine et al. 2017b, Data were then collected from base stations that were set up at multiple locations (see Fig. 1) on potential flyways (as identified during previous surveys and this study) to increase the chance of downloading data from birds flying past. Two base stations were set up in 2018 and four in 2019. The base stations automatically downloaded data from the tags when the bird flew in range (as per manufacturer's specifications, with an elevated base station and clear line of sight, this can occur many kilometers away). Base stations were powered using two car batteries (either West Marine 12V 32Ah, 20hr or Apex 12v 35Ah, 10hr) and a solar panel and remained powered for the full remaining breeding season in each of the two years. Base station locations were chosen by considering maximum line of sight for incoming birds and known flyways of birds based on previous observations. Base stations were deployed by teams in a helicopter, because the sites were too remote to hike to. Once gear was dropped in place, teams were deployed on foot to set the equipment up and ensure it was working.

Data analysis
Data were downloaded off SD cards and uploaded into Movebank (https://www.movebank.org/). Because data from the same tag was occasionally downloaded in part on one Base Station and in part on another (this happened multiple times), this program helped combine data from the same tag into a single track. Tracks were clipped to deployment date and time to remove any data from outside the deployment period. All data that achieved a positional fix and time stamp were then imported into R version 4.0.2 and ArcMap 10.8.1 (ESRI) for further analysis and mapping. For Tag 6507, which produced the vast majority of overland locational fixes, a kernel density estimate and 20% contour was produced using the "adehabitathr" package in R, using a smoothing factor determined with the ad hoc method. For all birds that returned flight information, distances flown were calculated using the "distHaversine" function in the R package "geo sphere" (Hijmans 2017). Ellipsoidal height values for each locational fix were corrected to orthometric height values (height above mean sea level) using the EGM2008 geoid model (Pavlis et al. 2012). We then used a 1/3 arc-second Digital Elevation Model from the USGS 3D Elevation Program (3DEP) Datasets from The National Map (USGS 2017), vertically referenced to height above local mean sea level, to calculate above ground heights for overland locational fixes. Ground heights above mean sea level were extracted to track points and subtracted from orthometric heights to obtain heights above ground surface. Individual locational fixes were identified as potential ground points if they had a height above ground of less than 10 m and a ground speed of less than 1 meter per second as recorded by the e-obs tags.

Burrow searching
Once data were downloaded and mapped, a burrow searching team was deployed to areas where tags showed the highest amount of concentrated circling activity and/or areas where the two birds landed on the ground. Because this was a remote area accessible only by helicopter, the number of ground searching trips that could be undertaken was limited by funding and logistics. Burrow searching was undertaken on two trips in 2018 (3 April to 4 May, and 16-20 July) and three trips in 2019 (6-9 Aug, 26-30 Aug, and 28-30 Oct). Teams searched the area identified by the tag data, concentrating on microhabitat features considered most suitable for Hawaiian Petrel burrows as identified by Raine et al. (2021) on Kauaʻi. This included searching for burrows on steep slopes with a high density of native plants characterized by ʻōhiʻa in the canopy and uluhe fern in the understory, with special attention paid to the immediate area around larger trees, around large tree roots, and mossy walls with small caves at the base. While searching these areas, the teams looked for any sign showing recent seabird activity, including guano, feathers, trampled trails, strong seabird scent, and, ultimately, active burrows.

RESULTS
A total of six tags were deployed, with two in 2018 and four in 2019 (Table 1). In 2018, one bird had a full brood patch and the second had one that was classified as partial. In 2019, two birds had brood patches classified as partial and two had no sign of a brood patch. Data were downloaded from four of six (66.7%) birds, but complete tracks were only recovered for two (33.3%) birds (#6507 and #7238). A total of 23,627 locational fixes were downloaded from tagged birds, with 1773 (7.5%) locational fixes recorded over land. The longest transmission period recorded from a single bird was 41 days.
In 2018, one bird (#6504) downloaded 379 points over one day (from the night it was tagged and early the following morning), with a single track going inland up the Waimanu Valley after tagging before the bird turned and headed out to sea. The second bird (#6507) downloaded 2497 points, which represented multiple short foraging trips up to 27.9 km offshore, and overland excursions and circling, for a total distance of 670 km traveled (Fig. 3A).
In 2019, two birds did not download any points and one downloaded 20 points as it departed the capture site. The fourth bird, (#7238) downloaded 20,731 points over a 41-day period, including a long foraging trip of 41 days and 14,800 km, and a single inbound overland trip 400 m up Waihlau Stream, a side drainage of the Waimanu Valley that included a ground location, with a following outbound overland trip (Fig. 3B). All overland tracks were interrogated in more detail (Figs. 4A and B) and the area of highest activity searched in both years. On 28 Oct 2019, the single ground location of the fourth bird was searched and an active Hawaiian Petrel burrow located (Fig. 4B).
Considering the GPS level accuracy of the tags, coupled with the apparently low density of breeding birds in the area, it was therefore assumed that this was the burrow of the tagged bird. Upon discovery the contents of the burrow were checked with a handheld camera and found to contain a Hawaiian Petrel chick. The burrow was monitored with a Reconyx Hyperfire HF2X covert camera for the rest of the season, but, unfortunately, because of an error in the camera settings, only took photos in the day so the chick fledging and exercising was missed (although the camera did capture images of a small Indian mongoose Herpestes javanicus, highlighting the threats of introduced predators in the area).
In that year, based on the presence of down around the burrow entrance and no sign of depredation, the chick was presumed to have fledged. The burrow was monitored with a Reconyx Hyperfire HP2X in 2020 and failed for an unknown reason, then monitored again with a camera in 2021 and fledged a chick (Fig. 5).

DISCUSSION
This study clearly demonstrated the utility of employing tracking devices to locate burrows of endangered Procellariforms, particularly those breeding in remote montane areas with challenging terrain. Using tracking devices attached to transiting Hawaiian Petrels, we successfully located the first documented Hawaiian Petrel burrow in the Puʻu O ʻUmi Natural Area Reserve (and the entirety of the Kohala Mountain), as well as multiple other areas of concentrated petrel activity.

Erratum:
In the original publication of this article an incorrect version of Figure 3 was published. The correct version of Figure 3 was put into place on 22 November 2022.

Fig. 5. Photo taken on 8 November 2021 by a Reconyx
Hyperfire HP2X camera of a Hawaiian Petrel (Pterodroma sandwichensis) chick exercising outside the burrow that was located by this study.

Tracking considerations
For this task we chose data loggers that automatically download data to static base stations; this was critical because there was little to no chance of recapturing tagged birds. e-obs tags were therefore used because of both their remote download capabilities and their high locational and altitudinal resolution (critical for narrowing down ground search areas). We also initially considered satellite tags and GSM transmitters for this purpose. Although data is automatically downloaded to satellite tags, they were discounted for two reasons. First, the spatial accuracy of satellite tags is lower than those of data loggers, meaning locational fixes would be coarser for burrow searching. Second, they do not provide locational fixes at the high frequency required for locating areas of concentrated activity required for locating burrows (10-20 points a day compared to GPS points every minute on the e-obs data loggers). GSM transmitters were discounted because of the limited cell coverage in the study area, meaning that data downloads would not be assured. Although the e-obs data loggers fit the criteria we needed for this study, they were relatively expensive (especially when factoring in the need for multiple base stations), which did limit the number of tags that could be deployed and thus the sample size.
The method used to catch birds for tagging (light attraction and playback) was also successful in catching transiting Hawaiian Petrels for tag attachment. Although we initially also utilized mist nets to potentially increase the chances of catching birds, these were found to be less useful than anticipated, and thus not used in subsequent capture sessions. It is worth noting, however, that the use of lights could also bias the sample to more sub-adults, because sub-adult seabirds are typically more attracted to lights than breeding birds (Reed et al. 1985, Le Corre et al. 2002, Rodríguez et al. 2017, although breeding adults are certainly attracted to bright lights at sea (Wiese et al. 2001, Merkel and Johansen 2011, Ronconi et al. 2015. If more sub-adults were tagged than breeding adults, this could explain why only two of the tagged birds downloaded complete tracks: sub-adults may not have returned to the area or may have flown over a wider area and never come into range of the base stations before the tags fell off. This is a more likely explanation than the possibility that the tags impacted the birds in a negative way, as the same tags and attachment methods have been used successfully on breeding Hawaiian Petrel adults on Kauaʻi and Lānaʻi with no impact on survival rates or fledging success from burrows with tagged parents (Raine et al. 2017b(Raine et al. , 2020b. To help ameliorate the potential sub-adult bias to this method, future tracking work should involve setting up the light wells adjacent to the valley where the burrow was ultimately found, because bright lights appear to be more likely to attract adult birds if they are immediately adjacent to the breeding colony (Raine, personal observation).
Placement of the base stations was also critical. We only had a small number of available base stations (two in the first year and four in the second), and these units were integral to collecting the data from tagged birds. If the birds did not fly within range of them, then no data would be downloaded. We placed base stations in the areas we thought had the most chance of being on inbound flight lines (based on observations by AW at Slant Camp, and tracking data collected from Hawaiian Petrels on Kauaʻi and Lānaʻi using the same method; Raine et al. 2017bRaine et al. , 2020b and had the farthest line of sight. Having a good understanding of terrain, wind patterns, and bird flight behavior is critical for studies such as these to ensure optimal placement and chance of success. Furthermore, ensuring that the solar panels of the base stations were placed in a way to maximize charging capabilities (in the open with a clear unobstructed view of the sky and at an angle to maximize full exposure to the sun) was also vitally important. Careful placement of our panels helped ensure that we did not have any issues charging the base stations even though the study area is often overcast, misty, and raining.
The complete tracks downloaded from the two individual birds also highlighted the potential differences in data produced using this method. Hawaiian Petrel #6507 demonstrated circling behavior over two distinct areas that occurred for several hours during the night, indicative of non-breeder courtship behavior (breeding adult Hawaiian Petrels do not engage in this behavior; Raine et al. 2017bRaine et al. , 2020b. On the other hand, Hawaiian Petrel #7238 undertook a long foraging trip (at 41 days this was much longer than the long foraging trips recorded from data loggers deployed on Hawaiian Petrels on the island of Lanai, which averaged 19.8d; Raine et al. 2020b) and then returned directly to what eventually turned out to be its burrow to feed its chick; therefore, clearly, a breeding adult. The two main areas with the most overland circling behavior (including the area with ground based points) by the presumed non-breeder was searched multiple times for signs of petrel breeding activity, and although both were found to be good breeding habitat for Hawaiian Petrel (Troy et al. 2017, no signs of ground activity were found. Although this does not mean there were no burrows present (as the terrain was very steep, much of it not searchable, and with dense understory), it is also possible that this area could be frequented by courting non-breeders further away from the actual breeding colony. Therefore, care needs to be taken in interpreting the tracking results collected from a study of this nature. On the other hand, the location of a single burrow does not necessarily mean this is where the majority of active burrows are (i.e., the located burrow could be in a low density breeding area), and so the data collected using the tags needs to be considered in conjunction with wider data available from acoustic recording devices and auditory surveys.

Conservation management implications
Identifying an active Hawaiian Petrel burrow and areas of concentrated breeding activity is a critical first step in initiating focused conservation actions to protect the endangered petrels nesting in this area. Whereas this species faces numerous threats at land and at sea, within breeding colonies primary threats include both loss of habitat from invasive plants (VanZandt et al. 2014) coupled with multiple introduced mammalian predators (including feral cats Felis catus, black rats Rattus rattus, feral pigs Sus scrofa, and small Indian mongoose [Simons 1985, Hodges and Nagata 2001, Hu et al. 2001, Raine et al. 2020a]), and the introduced Barn Owl Tyto alba . Addressing threats at colonies is logistically challenging and requires sufficient funding, so the results of this study allowed for intensive seabird management to be focused in areas where it is needed the most; around the located burrow and the highactivity areas identified from our analysis.
Subsequent to our study, extensive predator control trapping lines were created in the area and focused on cats, rats, and mongoose. This included five transects of 20 pairs of body-grip traps for cats, automatic self-resetting Goodnature A24 traps for rats (https:// goodnature.co/), and New Zealand Department of Conservation (DOC) 200 snap traps for mongoose, as well as monitoring cameras. A mongoose was recorded in 2019 on the camera placed at the burrow identified in this study, highlighting the constant threat facing the birds in this area. Funding was also recently acquired to build a small 0.6 hectare predator-proof fence in the vicinity, that will be coupled with the creation of a social attraction area (utilizing both a sound system to attract birds to the site and artificial burrows to enhance nesting potential; Miskelly and Taylor 2004, Buxton and Jones 2012, Gummer et al. 2014 to help increase the rate of colonization within this predator-free zone.
With an active burrow now located and management in place over the wider area, we are planning to increase the sample size of tracked birds from this study by deploying more tags in future years. This will be coupled with long-term acoustic monitoring using strategically placed acoustic sensors and multiple additional auditory surveys during peak breeding periods using night vision and thermal imaging. Combined, this work will help refine our understanding of the Hawaiian Petrel colonies of Puʻu O ʻUmi Natural Area Reserve in order to better protect them.
Responses to this article can be read online at: https://www.ace-eco.org/issues/responses.php/2328