Response of Australian Boobooks (Ninox boobook) to threatening processes across urban, agricultural, and woodland ecosystems Michael T. Lohr B.S. The Pennsylvania State University M.S. The University of Delaware Thesis Submitted for the degree of Doctor of Philosophy in the School of Science Edith Cowan University November 2019
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
Response of Australian Boobooks (Ninox boobook) to threatening
processes across urban, agricultural, and woodland ecosystems
Michael T. Lohr B.S. The Pennsylvania State University
M.S. The University of Delaware
Thesis Submitted for the degree of Doctor of Philosophy
in the School of Science
Edith Cowan University
November 2019
ii
“One of the penalties of an ecological education is that one lives alone in a world of wounds.
Much of the damage inflicted on land is quite invisible to laymen. An ecologist must either
harden his shell and make believe that the consequences of science are none of his
business, or he must be the doctor who sees the marks of death in a community that
believes itself well and does not want to be told otherwise.”
- Aldo Leopold, “A Sand County Almanac”
iii
Abstract The effects of habitat fragmentation on native wildlife can vary depending on the
type of land use occurring in the matrix between remaining habitat fragments. I used
Australian boobooks (Ninox boobook) in Western Australia to investigate interactions
between matrix type and four different potential threatening processes: secondary
poisoning by anticoagulant rodenticides (ARs); limitation of juvenile dispersal and impacts
on spatial genetic structure; breeding site availability; and infection by the parasite
Toxoplasma gondii.
I also conducted a literature review on the use and regulation of ARs in Australia and
published accounts of non-target impacts in order to contextualise exposure patterns
observed in boobooks. The review revealed records of confirmed or suspected poisoning
across 37 vertebrate species in Australia. World literature relating to AR exposure in
reptiles suggests that they may be less susceptible to AR poisoning than birds and mammals.
This relative resistance may create unevaluated risks for wildlife and humans in Australia
where reptiles are more abundant than in cooler regions where AR exposure has been
studied in greater depth.
I analysed AR residues in boobook livers across multiple habitat types. Second
generation anticoagulant rodenticides were detected in 72.6% of individuals sampled. Total
AR concentration correlated positively with the proportion of urban land use within an area
approximately the size of a boobook’s home range centred on the point where the sample
was collected. ARs originating in urban habitat probably pose a substantial threat to
boobooks and other predatory wildlife species.
No spatial genetic structure was evident in boobooks across habitat types. I
observed one individual dispersing at least 26km from its natal home range across urban
habitat. The apparent permeability of anthropogenically altered landscapes probably
explains the lack of spatial genetic structure and is likely related to the observed ability of
boobooks to use resources in both urban and agricultural matrices.
Boobooks did not appear to be limited by the availability of suitable nesting sites in
urban or agricultural landscapes. Occupancy did not change significantly over the duration
of the study in remnants provided with artificial nest boxes in either landscape type.
iv
However, in one instance, boobooks successfully used a nest box located in an urban
bushland. Nest boxes may be a useful management tool in highly-altered areas where
natural hollows are unavailable.
Toxoplasma gondii seropositivity in boobooks did not vary significantly by landscape
type but was more prevalent in individuals sampled during cooler wetter times of year. Risk
of exposure due to greater cat abundance in urban and agricultural landscapes may be
offset by creation of environmental conditions less favourable to the survival of T. gondii
oocysts in soil.
Taken together, this body of research demonstrates variation in relationships
between different types of habitat fragmentation and threatening processes related to
fragmentation. This research also raises questions about how habitat fragmentation is
discussed and studied in the context of species which are capable of making extensive use
of matrix habitat. I recommend greater consideration of the concept of “usable space”
when studying fragmentation impacts in habitat generalists.
v
Declaration
I certify that this thesis does not, to the best of my knowledge and belief:
i. incorporate without acknowledgment any material previously submitted for a degree or
diploma in any institution of higher education;
ii. contain any material previously published or written by another person except where
due reference is made in the text; or
iii. contain any defamatory material.
iv. I also grant permission for the Library at Edith Cowan University to make duplicate
copies of my thesis as required.
Michael T. Lohr
06/11/2019
vi
Acknowledgments
I would first like to thank my supervisors Dr. Rob Davis and Dr. Allan Burbidge. Their
insights into navigating the complex ecosystem that is conservation research in Western
Australia are greatly appreciated. I sincerely appreciate the free rein they gave me in
exploring a series of sometimes unconventional side projects. These opportunities have
proved invaluable. Rob’s willingness to meet at length to discuss new opportunities and
troubleshoot occasional difficulties made the entire PhD experience easier and more
enjoyable.
Cheryl, your accommodation of my bizarre nocturnal field schedule, financial
support, tolerance for endless monologues about anticoagulant rodenticides, and R code
are what made this whole thing actually work. Thank you. I look forward to having our life
back in the near future.
Many thanks to the large number of people and organisations willing to hold their
collective noses and accumulate dead owls for me. This PhD would not have been possible
without your efforts. I hope to continue to do my part to convert the smelly data you
collected into meaningful conservation actions. Samples were contributed by Kanyana
Wildlife Rehabilitation, Native Animal Rescue, Native ARC, Nature Conservation Margaret
River Region, Eagles Heritage Wildlife Centre, and many individual volunteers especially
Steve Castan, Simon Cherriman, Angela Febey, Warren Goodwin, Amanda Payne, Stuart
Payne, and Boyd Wykes.
Many people provided help on long nights of owl surveys and nest box checks
Australian Ringneck (Barnardius zonarius) N/A pindone suspected Western Australia rabbit control primary herbivore Twigg et al., 1999 Brahminy Kite (Haliastur indus) N/A pindone suspected Western Australia rabbit control secondary carnivore Twigg et al., 1999 Crested Pigeon (Ocyphaps lophotes) N/A pindone known* Western Australia rabbit control primary herbivore Twigg et al., 1999 Grass Owl (Tyto longimembris) 1 brodifacoum liver analysis Queensland
agricultural rat control secondary carnivore Young & De Lai, 1997
Masked Owl (Tyto novaehollandiae) 1 brodifacoum
physical symptoms Queensland
agricultural rat control secondary carnivore Young & De Lai, 1997
Rufous Owl (Ninox rufa) 2 brodifacoum
physical symptoms Queensland
agricultural rat control secondary carnivore Young & De Lai, 1997
Mammals southern brown
bandicoots (Isoodon obesulus) N/A pindone liver analysis Western Australia rabbit control primary omnivore Twigg et al., 1999 swamp wallaby (Wallabia bicolor) N/A pindone known* New South Wales rabbit control primary herbivore Twigg et al., 1999 western grey kangaroo (Macropus fuliginosus) N/A pindone known* Western Australia rabbit control primary herbivore Twigg et al., 1999 brushtail possum (Trichosurus vulpecula) 7 unknown
physical symptoms Queensland unknown
not specified omnivore Grillo et al., 2016
boodie (Bettongia lesueur) 20-50 pindone
population eradicated Western Australia
island rat eradication primary herbivore Morris, 2002
20
In addition to accounts of wildlife poisoning, we also located published accounts 2
suggesting population-level effects of rodenticide toxicity on carnivorous birds in Australia. 3
Olsen (1996) listed the use of rodenticides in areas of palm cultivation as a potential 4
contributing factor in the decline of Norfolk Island Boobooks (Ninox novaeseelandiae 5
undulata x novaeseelandiae). Young and De Lai (1997) observed a correlation between 6
declines in owl abundance and the use of “Klerat®”a brodifacoum-based rodenticide in 7
sugar cane fields in north Queensland and documented one confirmed and several 8
suspected cases of brodifacoum poisoning in owls (James, 1997). A subsequent report 9
noted three additional cases of owls in Queensland testing positive for brodifacoum 10
residues (0.007 mg/kg, <0.005mg/kg, and 0.17mg/kg ) in the 1990s and two museum 11
specimens of Southern Boobooks (Ninox novaeseelandiae) with rodenticide poisoning listed 12
as their cause of death in the collection notes (Thomas and Kutt, 1997). One of the two 13
specimens, while alive showed symptoms of AR poisoning including “bleeding from the 14
nasal passages; loss of muscle co-ordination; lethargy including drooping head and eyes; 15
and generally poor and dirty condition” (Thomas and Kutt, 1997). The report reviewed 16
several other factors which could potentially have impacted owl populations in the area and 17
came to the conclusion that there was “significant potential for secondary poisoning of owls 18
to occur in Queensland sugarcane as a result of the use of Klerat®” (Thomas and Kutt, 1997). 19
Crop Care Australia later deregistered Klerat® for use in sugar cane fields over concerns 20
relating to secondary poisoning (Twigg et al., 1999). 21
An unpublished PhD dissertation examined dynamics of secondary poisoning of 22
avian predators associated with sugar cane fields in Queensland and concluded that the 23
coumatetralyl-based product used to control rats did not pose a threat to predatory birds 24
(Ward, 2008). This conclusion was based largely on the low relative use of canefields for 25
foraging by predatory birds, the low concentration of coumatetralyl in rats captured outside 26
of canefields, and the low toxicity and persistence of coumatetralyl relative to second 27
generation anticoagulant rodenticides (Ward, 2008). Unfortunately, no predatory birds in 28
the treated areas were directly tested for rodenticide exposure. A lack of detection of 29
coumatetralyl in Southern Boobooks in Western Australia as part of an ongoing study 30
supports the low probability of secondary toxicity in raptors for this rodenticide. 31
21
Pindone has been implicated as a factor driving the decline of Little Eagle (Hieraaetus 32
morphnoides) numbers in and around Canberra (Olsen et al. 2013). Breeding pairs of Little 33
Eagles disappeared from areas baited with pindone while pairs in areas baited with 1080 or 34
not baited at all persisted (Olsen et al. 2013). The high susceptibility of Wedge-tailed Eagles 35
to pindone in laboratory tests (Martin et al. 1994) lends credibility to the hypothesis that 36
pindone could be responsible. Unfortunately, no direct testing of Little Eagles suspected of 37
being poisoned was conducted to confirm pindone exposure and rule out other ARs from 38
residential and commercial sources. 39
Recently, a study in Tasmania examined probable rodenticide poisoning in predatory 40
birds. Six species (Table 2.2) showed signs of anticoagulant rodenticide poisoning when 41
dissected (Mooney, 2017) but the rodenticides responsible were not determined or 42
quantified. As part of this study, thirteen predatory bird species were ranked by risk of 43
rodenticide exposure according to four natural history parameters: relative metabolic 44
for rodents, and willingness to forage near anthropogenic structures (Mooney, 2017). 46
Development of a more statistically robust predictive model using similar natural history 47
parameters to examine risk of rodenticide exposure in a wider range of predatory species 48
would be an extremely useful step toward assessing likely population level impacts on 49
wildlife in Australia. Incorporating variables relating to seasonal dietary shifts and home 50
range size could potentially improve future models. 51
The overall lack of attention within Australia to what is perceived as a potentially 52
serious threatening process for native carnivores in many other parts of the world suggests 53
the need for Australian studies which examine potential impacts on native fauna in a 54
quantitative and comprehensive manner. Susceptibility of marsupial carnivores is 55
particularly poorly understood and should be a focus of future research. Furthermore, a 56
surveillance program should be in place in areas of high AR use, to monitor any dead wildlife 57
for a cause of death. Most of the studies we used in this review did not sample animals and 58
thus were not able to confirm suspicions of death due to rodenticide poisoning. 59
4.3 Governance and legislation of rodenticide use 60
22
At present, no information is available on the volume of sales or application of ARs in 61
Australia. Reporting for all poisons intended to control vertebrates indicates that 222 62
different products are currently registered with a total sales reaching $18,601,875.00 in the 63
2015-2016 fiscal year (Australian Pesticides and Veterinary Medicines Authority, 2017a). 64
Nine anticoagulants are currently approved for vertebrate pest control in Australia (McLeod 65
and Saunders 2013). At present, all nine are listed as Schedule 6 substances (see Appendix 66
2.A for schedule meanings) in Australia (Australian Government Department of Health: 67
Therapeutic Goods Administration, 2017) (Table 2.3) and are legally allowed to be sold 68
directly to the public and do not require government permits for purchase or use. In some 69
cases, more concentrated formulations of SGARs are listed as Schedule 7 substances and are 70
restricted to licensed pesticide applicators (Australian Government Department of Health: 71
Therapeutic Goods Administration, 2017) while products containing low concentrations of 72
some FGARs are registered as schedule 5 substances which require only simple warnings 73
and safety directions for public sale (Table 2.3). The FGAR diphacinone is currently approved 74
as an active ingredient but has no products registered with the APVMA after July 2016 75
(Australian Pesticides and Veterinary Medicines Authority, 2017b). However, remaining 76
stock can still be used for 12 months following a stopped registration (Commonwealth of 77
Australia, 1994) and MSDS sheets obtained from a pest management contractor seem to 78
indicate that at least one diphacinone product is still in use at present. The APVMA has 79
prioritised a review of the status of all SGARs currently approved in Australia (brodifacoum, 80
bromadiolone, difenacoum, difethialone, and flocoumafen) citing concerns over public 81
health, worker safety, and environmental safety (Australian Pesticides and Veterinary 82
Medicines Authority, 2015)83
23
Table 2.3 Anticoagulants currently approved for vertebrate pest control in Australia. Some anticoagulants are assigned different schedules dependant on formulation. *Some disagreement 84 exists as to whether these should be treated as first or second generation anticoagulants †Warfarin is used therapeutically in humans as a blood thinner. 85
Anticoagulant Chemical Class Generation Schedule (See Appendix 2.A)
Acute Oral LD50 (Rattus norvegicus)
mg/kg
LD50 Reference Approved
Target Species
brodifacoum hydroxycoumarins second 6 (0.25 per cent or less) or 7 0.27
Godfrey 1985 mice and rats
bromadiolone hydroxycoumarins second 6 (0.25 per cent or less)or 7
0.57-0.75 Meehan 1978 mice and rats
coumatetralyl hydroxycoumarins first 5 ( 0.05 per cent or less), 6 (1 per cent or less), or 7 16.5
Dubock and Kaukeinen 1978 mice and rats
difenacoum hydroxycoumarins second 6 (0.25 per cent or less) or 7 1.8-3.5
recycling of vitamin K in the liver, which subsequently disrupts normal blood clotting in 639
vertebrates (Park et al. 1984). ARs are often divided into first generation anticoagulant 640
rodenticides (FGARs) and second generation anticoagulant rodenticides (SGARs) based on 641
their chemical structure and when they were first synthesized. Unlike FGARS, SGARs are 642
often lethal with a single feed and are substantially more persistent in liver tissue (Erickson 643
and Urban, 2004). 644
AR exposure and subsequent mortality have been detected in non-target wildlife in 645
all parts of the world where exposure has been tested (Laakso et al., 2010). Predatory bird 646
species are particularly vulnerable to AR poisoning due to a greater susceptibility to most 647
ARs than other bird species (Herring et al., 2017) and a prey base which frequently contains 648
rodents targeted by the use of ARs. In some raptor species, mortality from AR exposure 649
may have population-level impacts (Thomas et al., 2011). Unlike in Europe and North 650
America, where the non-target impacts of ARs have been extensively studied, relatively 651
little research has been conducted on AR exposure in Australian wildlife (Lohr and Davis, 652
2018; Olsen et al., 2013). This knowledge gap exists despite several lines of evidence 653
suggesting that patterns of regulation and usage in combination with differences in faunal 654
assemblages may increase the incidence and severity of non-target AR poisoning in Australia 655
relative to better-studied areas of the world (Lohr and Davis, 2018). 656
Within Australia, patterns in the spatial distribution of AR exposure have not been 657
studied in any wildlife species. A number of studies have addressed the spatial ecology of 658
anticoagulant rodenticide exposure in non-target wildlife but have been primarily limited to 659
North American mammals. Of these, some have focused on impacts within specific habitat 660
types (Cypher et al., 2014; Gabriel et al., 2012). Studies examining patterns of AR exposure 661
between urban and rural habitats have found correlations between the use of urban habitat 662
and exposure rates in San Joaquin kit foxes (Mcmillin et al., 2008) and bobcats (Riley et al., 663
2007). A model developed to predict exposure patterns in San Joaquin kit foxes found that 664
exposure was most likely in areas of low density housing on the urban/rural interface 665
(Nogeire et al., 2015). Similar dynamics have been suggested but not tested in predatory 666
bird species. Studies in North America and Europe have noted that predatory bird species 667
which use more developed habitats tend to have greater rates of AR exposure than those 668
which predominantly use more natural landscapes (Albert et al., 2010; Christensen et al., 669
2012). Additionally, a study in Spain noted a positive correlation between human 670
44
population density and AR exposure in a sample of 11 species of predatory birds and 671
mammals (López-Perea et al., 2015). The greater use of rodenticides and higher prevalence 672
of targeted commensal rodents in human-dominated landscapes relative to natural areas is 673
likely to drive these observed and suggested differences in non-target exposure. However, 674
because AR usage patterns differ between urban and agricultural environments (Lohr and 675
Davis, 2018) a need exists to evaluate the possibility of differences in non-target exposure 676
patterns between different types of anthropogenic landscapes. 677
To address this knowledge gap, I sought to compare anticoagulant rodenticide (AR) 678
exposure across intact native bushland and two different types of anthropogenic 679
landscapes. Additionally, I undertook the first large-scale targeted testing of wildlife for AR 680
exposure in the continent of Australia (Lohr and Davis, 2018). Testing was conducted on 681
Southern Boobooks (Ninox boobook), which provide an excellent model to quantify the 682
spatial distribution of threatening processes associated with fragmentation due to their 683
presence across multiple habitat types and high abundance relative to other predatory bird 684
species. To the best of my knowledge, no studies have directly addressed the relative 685
impacts of different types of human land use on AR exposure in non-target wildlife. 686
Understanding how different types of human land use impact the likelihood of AR exposure 687
in non-target wildlife will be critical in evaluating risks to wildlife on a continental scale and 688
will enable more effective targeting of measures to mitigate secondary toxicity. 689
Methods 690
Southern Boobooks are medium-sized hawk owls found across the majority of 691
mainland Australia and adjacent parts of Indonesia and New Guinea (Olsen, 2011a). They 692
are assigned a conservation status of “Least Concern” by the IUCN (“Ninox boobook,” 2018). 693
Some taxonomies consider Southern Boobooks to be synonymous with the closely-related 694
New Zealand Morepork (Ninox novaseelandiae) found in Tasmania and New Zealand but 695
recent genetic and bioacoustic evidence suggests otherwise (Gwee et al., 2017). Boobooks 696
are dietary generalists, consuming a wide variety of vertebrate and invertebrate prey 697
(Higgins, 1999; Trost et al., 2008). These dietary habits make them an ideal model species 698
for broad assessment of contamination of food webs by persistent pollutants like ARs. Their 699
presence in most habitat types across Australia, with the exception of treeless deserts 700
45
(Higgins, 1999), facilitates examination of differences in exposure across multiple habitat 701
types and allows for future replication of this study at sites across the continent. 702
Specimen Collection 703
Dead boobooks found in Western Australia were solicited from a network of 704
volunteers, wildlife care centres, and government departments and were opportunistically 705
collected when encountered. Boobooks euthanized by veterinarians and wildlife 706
rehabilitators due to severe disease or injury were included. Dates and locations where 707
each boobook was initially collected were recorded from the collector when possible. If 708
liver tissue was identifiable and had a mass >3g, it was removed and stored frozen at 20°C 709
until analysed for AR residues. A total of 73 usable boobook livers were stored for testing. 710
While an effort was made to obtain boobooks from a diversity of geographical areas and 711
habitat types throughout Western Australia, most samples originated in the more densely 712
settled urban and peri-urban areas in the south-west of Western Australia in and around the 713
city of Perth. 714
Rodenticide Analysis 715
Liver samples were analysed by the National Measurement Institute (Melbourne, 716
Australia) for residues of three FGARs (warfarin, coumatetralyl, and pindone) and five SGARs 717
(difenacoum, bromadiolone, brodifacoum, difethialone, and flocoumafen) registered for use 718
in Australia by the Australian Pesticides and Veterinary Medicines Authority. For each 719
sample, 10ml of reverse osmosis water and one gram of liver tissue were added to a 50ml 720
analytical tube and shaken for 15 minutes on a horizontal shaker. A 10ml volume of 5% 721
formic acid in acetonitrile solution was then added and the tube was shaken for an 722
additional 30 minutes. QuEChERS extraction salt was added and the tube was shaken for an 723
additional two minutes. The tube was then centrifuged for 10 minutes at 5100rpm. After 724
pipetting 3ml of the supernatant into a 15 ml analytical tube, 5ml of hexane was added and 725
the tube was shaken for two minutes then centrifuged for 10mins at 5100rpm. The hexane 726
layer was removed using a vacuum pipette and discarded. A 1ml aliquot of the supernatant 727
was transferred to a 2ml QuEChERS dispersive tube, shaken for one minute, and centrifuged 728
at 13000rpm for three minutes. The QuEChERS supernatant was then filtered using a 729
0.45μm filter. After filtration, 3μl of coumachlor was added as an internal standard to 497μl 730
of the filtered extract and vortexed prior to LC-MS/MS analysis. A Waters TQS Tandem 731
46
Quadrupole Detector Liquid Chromatograph-Mass Spectrometer (LC-MS/MS) and an 732
Acquity UPLC CSH C18 100 x 2.1mm column were used to quantify concentrations of each 733
rodenticide. Recovery rates for each AR, were calculated using chicken liver samples spiked 734
with analytical standards (Table 3.1). 735
Table 3.1 Limit of detection (LOD), limit of quantification (LOQ), average recovery, and relative standard deviation (RSD) for 736 eight ARs in a spiked chicken liver matrix. 737
Compound LOD (mg/kg) LOQ (mg/kg) Average recovery % (RSD)
Warfarin 0.001 0.002 94 (8.1)
Coumatetralyl 0.001 0.002 93 (7.6)
Bromadiolone 0.005 0.010 96 (9.5)
Difenacoum 0.005 0.010 96 (11.2)
Flocoumafen 0.005 0.010 103 (11.4)
Brodifacoum 0.005 0.010 92 (8.8)
Difethialone 0.005 0.010 91 (14.6)
Pindone 0.005 0.010 36 (13.5)
738
Statistical Analysis 739
Total AR liver concentration is commonly used to compare toxicity risk when 740
individuals are exposed to multiple rodenticides (Christensen et al., 2012) due to similarities 741
in their modes of action and likely cumulative effects (Hughes et al., 2013). For this reason, 742
the sum of all liver rodenticide concentrations above the limit of detection was calculated 743
for each individual for the purposes of comparing differences in exposure by age, season, 744
and land use. In order to compare seasonal trends in total AR concentration, boobooks 745
were assigned to four groups based on their collection date: summer (December –746
February), autumn (March-May), winter (June- August), and spring (September-November). 747
All boobooks with known collection months (n=71) were included in the seasonal analysis. 748
The Kruskal-Wallis test was used to assess whether significant differences existed in liver AR 749
concentration by season. 750
Boobooks were assigned to age classes of less than one year ("hatch year") or 751
greater than one year ("after hatch year") based on the presence of juvenile down and by 752
examination of fluorescence patterns under ultraviolet light (Weidensaul et al., 2011). In 753
one instance, it was not possible to determine age class due to degradation of porphyrins 754
caused by prolonged exposure of ventral remiges to sunlight. A total of 72 boobooks of 755
determined age class were available for analysis of the relationship between age and AR 756
exposure. I used a Mann-Whitney-Wilcoxon test to determine whether total liver 757
47
concentration of ARs varied between the two age classes. Results were considered 758
significant if p<0.05. 759
Exposure Thresholds 760
The utility of rodenticide concentration in liver tissue as a means to diagnose lethal 761
exposure has been questioned (Erickson and Urban, 2004; Thomas et al., 2011) as 762
susceptibility to acute toxicity can vary among individuals and across species (Thomas et al., 763
2011). Exposure to multiple ARs adds additional complexity to the assessment of likely 764
impacts from residual liver concentrations (Murray, 2017). However, a need exists to 765
estimate likely impacts across exposed individuals and to compare the magnitude of 766
exposure to previous studies. Accordingly, I identified relevant literature which established 767
commonly used guidelines for outcomes of various exposure rates in related taxa to allow 768
estimation of likely impacts on boobooks. 769
The Rodenticide Registrants Task Force suggested that a 0.7 mg/kg liver 770
concentration of brodifacoum was likely to be toxic based largely on captive studies of Barn 771
Owls (Kaukeinen et al., 2000), however this threshold estimate may be too high, as 772
environmental conditions affecting wild birds may increase their susceptibility to ARs 773
relative to captive birds (Mendenhall and Pank, 1980). Dowding et al. (1999) estimated a 774
lethal liver concentration for brodifacoum of 0.5 mg/kg using 29 individuals from 10 species 775
of birds. Numerous studies have reported thresholds of 0.2 mg/kg (Albert et al., 2010; 776
Christensen et al., 2012; Hughes et al., 2013; Langford et al., 2013; López-Perea et al., 2015; 777
Stansley et al., 2014; Walker et al., 2008) and 0.1 mg/kg (Albert et al., 2010; Christensen et 778
al., 2012; Langford et al., 2013; Ruiz-Suárez et al., 2014; Shore et al., 2016; Stansley et al., 779
2014; Walker et al., 2011, 2008) as indices of lower limits at which lethal AR toxicity was 780
likely to occur in predatory birds. These estimates were based on two studies examining 781
wild barn owls: Newton et al. (1999) and Newton et al. (1998) respectively. I also included a 782
threshold of 0.01mg/kg as this is the lowest published record of lethal SGAR toxicity in a 783
predatory bird species (Stone et al., 1999). Boobook liver concentrations were compared 784
against these thresholds (0.7 mg/kg, 0.5mg/kg, 0.2 mg/kg, 0.1 mg/kg, and 0.01mg/kg) to 785
facilitate a comprehensive understanding of overall potential impacts of ARs across all 786
sampled individuals. 787
48
Spatial Analysis 788
Only boobooks with accurate location data were included in the spatial analysis. In 789
one instance, two road-killed boobooks were recovered at the same location. One of these 790
was randomly removed from the spatial analysis, leaving a total of 66 boobooks available 791
for analysis. Land cover for the state of WA was classified into developed, agriculture, 792
native vegetation or open water. The developed category included all areas with 793
anthropogenic impervious surfaces (roads, buildings car parks, etc.) as well as intensive land 794
uses that did not qualify as agriculture (mines, landfills, sports grounds, golf courses etc.). 795
The agriculture category included a diversity of irrigated and dryland crops, orchards, and 796
grazed areas. Intensive indoor animal agriculture was included in the developed category 797
rather than agriculture because it consisted primarily of buildings and other impervious 798
surfaces. Areas subjected to silvicultural practices were classified as part of the native 799
vegetation category due to structural similarity. Additionally like native bushland, the only 800
anticoagulant permitted for use in forestry is pindone which is used to control rabbits in 801
areas too close to human habitation to allow the safe use of 1080. Percentages of each 802
classification were calculated within circular buffer zones (areas of influence) of three 803
different sizes around each location where a boobook was found. The two smaller buffer 804
sizes were calculated to match the mean area of a boobook’s core home range (7.3 ha) and 805
total home range (145.1 ha) (Olsen et al., 2011). The largest buffer size was an arbitrarily 806
large area with a 3km radius (2827.4 ha). This larger buffer was included to account for the 807
possibility of movement of contaminated prey into boobooks’ home ranges from adjacent 808
areas influencing the probability of boobook exposure to ARs. Because open water was not 809
considered to be usable space, the percentages of the other three habitat types were 810
calculated excluding any open water within the buffers. 811
I used general linear models with a negative binomial distribution, following 812
methodology used by Christensen et al. (2012), to analyse differences in rodenticide 813
exposure by habitat composition at the three different spatial scales. The Akaike 814
Information Criterion AIC was used to rank models for habitat proportions at each spatial 815
scale. Only single variable models were considered in the ranking due to nesting and 816
correlation of habitat proportions and spatial scales. I calculated McFadden's pseudo-R2 817
values for each habitat type and spatial scale combination. Statistical analysis was 818
performed using RStudio 1.1.383 (RStudio, Inc., Boston, MA, USA). 819
49
Results 820
While I did not directly quantify physiological signs of rodenticide poisoning due to 821
most carcasses being damaged as a result of vehicle collisions, during dissection I observed 822
symptoms associated with acute lethal AR toxicity in at least nine boobooks exhibiting no 823
sign of trauma. These symptoms included excessive bleeding from minor lacerations, pale 824
or mottled livers, subdermal and muscular haemorrhage in the absence of trauma, blood in 825
the thoracic cavity, and blood around the mouth and nares. Similar symptoms have been 826
described in association with lethal AR toxicity in other raptor species (Murray, 2017). 827
50
Table 3.2 Percentage exposure, mean exposure and total detection of eight different anticoagulant rodenticides in livers of 73 Southern Boobooks in Western Australia. 828
Coumatetralyl Warfarin Pindone Difenacoum Brodifacoum Bromadiolone Difethialone Flocoumafen Total
Table 3.3 Published rates of multiple second generation anticoagulant rodenticide exposure and percentages of individuals with exposure above two thresholds in predatory birds. 840
Species Location n Individuals % Exposed % Multiple Exposure
% >0.1 mg/kg
% >0.2 mg/kg
Mean Exposure (mg/kg) (SE)
Source
Southern Boobook (Ninox boobook) Western Australia 73 72.6 38.4 50.7 35.6 0.310 (0.069) this study
Tawny Owl (Strix aluco) United Kingdom 172 19.2 2.9 12.2 5.8 0.125 Walker et al., 2008
ARs were detected in 72.6% of all boobook liver samples (Table 3.2) with a mean 841
summed AR exposure of 0.310 mg/kg (SE 0.069246735) (Table 3.3). Approximately 17.8% of 842
boobook livers contained greater than the suspected lethal threshold of 0.5 mg/kg total ARs 843
(Figure 3.1) with 13.7% above the more conservative limit of 0.7 mg/kg. Seven of the ten 844
boobooks with AR liver concentrations above 0.7 mg/kg appear to have died directly of AR 845
poisoning and the other three showed signs of poisoning described by Murray (2017) 846
despite other apparent proximate causes of death. More than half of the boobooks tested 847
had liver concentrations above 0.1 mg/kg (Figure 3.1) and would likely have experienced at 848
least some degree of coagulopathy (Rattner et al., 2014a). The majority of boobooks 849
(65.8%) were exposed at a level above 0.01 mg/kg – the lowest observed lethal threshold in 850
an owl (Figure 3.1). 851
852
Figure 3.1 Percentages of Southern Boobooks (n=73) in Western Australia exposed to rodenticides stratified by total 853 rodenticide liver concentration (mg/kg) thresholds indicating potential outcomes. 854
The three FGARs tested – coumatetralyl, warfarin, and pindone – were infrequently 855
detected and accounted for only 0.01% of all ARs detected (Table 3.2). Coumatetralyl and 856
pindone were not detected in any of the samples and warfarin was detected in two 857
individuals at low levels (0.0024 mg/Kg and 0.0014 mg/Kg). The lower of these was below 858
the limit of quantification. Detectable exposure to SGARs was substantially higher (Table 859
3.2). Brodifacoum – the most commonly detected SGAR – was found in 72.6% of samples 860
55
and made up 84.0% of all rodenticides detected by mg/kg. It was detected in all liver 861
samples containing AR residues (Table 3.2). Difethialone and flocoumafen, which were not 862
known to be in use by the public were also detected in boobooks. Two or more ARs were 863
detected in 38.4% of boobooks tested (Figure 3.2). A maximum of five different ARs was 864
detected in two individual boobooks. 865
866
Figure 3.2 Percentages of Southern Boobooks (n = 73) exposed to multiple anticoagulant rodenticides in Western Australia. 867
Mean total liver concentration of ARs was not significantly different between age 868
classes (p= 0.34). AR exposure was greatest in boobooks collected in winter and winter 869
concentrations were significantly different from summer concentrations (p=0.026) (Figure 870
3.3). The livers of two recent fledglings still under parental care contained low but 871
quantifiable amounts of brodifacoum (0.022 and 0.051 mg/kg) and difethialone (0.020 and 872
0.022 mg/kg). 873
56
874
Figure 3.3 Mean total anticoagulant rodenticide concentration (mg/kg) in liver tissue of Southern Boobooks (n= 71) in 875 Western Australia by season. 876
Total AR exposure was positively correlated with the amount of developed area 877
within buffers at all spatial scales (Table 3.4). Proportions of agriculture and bushland 878
habitat within buffers were negatively correlated with total AR exposure at all spatial scales 879
(Table 3.4). The three AIC top-ranked models quantified habitat composition at the scale of 880
a full boobook home range and were all statistically significant (Table 3.4). The top-ranked 881
model used developed habitat at the scale of a boobook’s total home range and was highly 882
significant (p=0.00182). Correlations between the top three ranked models and total AR 883
57
concentration were not particularly strong but are stronger than would be suggested by 884
interpretation of traditional R2 indices, as McFadden's pseudo-R2 values falling in the range 885
of 0.2 to 0.4 “represent an excellent fit” (McFadden, 1978). 886
Table 3.4 Akaike information criterion (AIC) ranking of models of the association between percentage of single land use 887 types within buffers around collection points and total anticoagulant rodenticide liver concentration in Southern Boobooks 888 (n= 66) in Western Australia at three different spatial scales (Big=2827.4 ha buffer, Mid=145.1 ha buffer, Small=7.3 ha 889 buffer. 890
Model Estimate Std. Error z value Pr(>|z|) AIC McFadden's pseudo-R2
Mid Developed 2.1439 0.6876 3.118 0.00182 751.43 0.08675021
novaehollandiae) (Olsen et al., 1990), Brown Goshawks (Accipiter fasciatus) (Czechura et al. 1424
1987), and larger owls (Debus, 2009). Therefore, we hypothesised that they would be less 1425
likely to cross large sparsely-vegetated agricultural areas where they could be exposed to 1426
greater predation risk. That is, the matrix could be considered more hostile and less 1427
permeable in agricultural regions than in urban areas. Under these circumstances, 1428
fragmentation by agriculture in the wheatbelt could be functionally different to urban 1429
fragmentation in Perth with regard to dispersal and subsequent genetic impacts. 1430
Determination of genetic connectivity in boobooks via genetic analysis of individuals 1431
on a landscape scale will help settle long-standing speculation about the basic biology of this 1432
species and inform the management of an ecologically important and widespread avian 1433
carnivore. We aimed to determine whether potential differences in permeability of 1434
different types of anthropogenically-altered landscapes impacted genetic diversity and gene 1435
flow in a common but declining predatory bird by examining patterns of spatial genetic 1436
structure and corroborating these data with movement data derived from mark-recapture 1437
studies. We predicted that barriers to gene flow would occur in both urban and agricultural 1438
landscapes but would be more apparent in habitat fragmented by agricultural land use due 1439
to reduced dispersal capacity across a more hostile matrix. 1440
76
Methods 1441
Juvenile Dispersal 1442
To directly assess boobook dispersal capacity across fragmented habitats, we captured 1443
boobooks as nestlings or recent fledglings within their natal territory. Each young boobook 1444
was fitted with an individually-numbered stainless steel band issued by the Australian Bird 1445
and Bat Banding Scheme (ABBBS) to allow subsequent identification if re-sighted alive or 1446
recovered dead. A total of 17 boobooks from seven family groups were captured and 1447
banded. Of these, five individuals were re-sighted or found dead elsewhere. Location data 1448
submitted to the ABBBS by members of the public were then used to estimate dispersal 1449
distances. We also accessed data from the ABBBS from other banding studies elsewhere in 1450
Australia. We only included records of healthy birds captured in the wild to avoid potential 1451
bias from records of rehabilitated birds which may have behaved abnormally or been 1452
released away from the location where they were found. We found only eight additional 1453
qualifying instances of boobooks in Australia being banded as juveniles or nestlings and 1454
subsequently being resighted. One of these records was removed because the boobook 1455
was later recovered dead and still in the nest, leaving 12 available records, including those 1456
generated by our study. 1457
Genetic Sample Collection 1458
Western Australia is the largest state in Australia and covers an area of 1459
approximately 2,529,875 km² and makes up roughly the western third of the continent of 1460
Australia. We opportunistically collected blood and tissue samples from across the entirety 1461
of the state (Figure 4.1). We attempted to focus collection effort on three areas: the Perth 1462
metropolitan area, adjacent areas of continuous bushland in the Perth Hills, and agricultural 1463
areas in the agricultural Wheatbelt region in the vicinity the town of Kellerberrin 1464
approximately 200km east of Perth, in order to examine genetic structure across three 1465
distinct habitat types. 1466
77
1467
Figure 4.1 Sample locations of genotyped Australian Boobooks (Ninox boobook) in Western Australia. (“metro” = urban and 1468 suburban areas of Perth represented by squares, “rural” = forested area surrounding the Perth Metropolitan area 1469 represented by an “x” , “Southwest WA” = forested areas to the south of Perth represented by triangles, “Wheatbelt” = 1470 highly-fragmented agricultural landscapes represented by crosses, and “other” = Goldfields and Pilbara regions, 1471 represented by black circles, ‘other’ = Goldfields and Pilbara regions of Western Australia). 1472
We used several methods to collect genetic information. Live Australian Boobooks 1473
were captured using a noose pole (Olsen et al. 2011) at night in conjunction with audio lures 1474
while conducting occupancy surveys across landscapes dominated by urban, bushland, and 1475
agricultural habitats. Additional wild boobooks were captured opportunistically using a 1476
noose pole when roosting individuals and family groups were reported by volunteers during 1477
the day. Blood was also collected from live boobooks held by wildlife rehabilitators along 1478
with information about where the boobook was originally found. Blood was drawn from the 1479
right jugular vein of each captured boobook using an insulin syringe with a 25G needle 1480
designed for subcutaneous use. In larger birds where more than a single capillary tube of 1481
blood is required, it is preferable to take blood from the right jugular vein, as this reduces 1482
handling time and risk of hematoma relative to sampling from the brachial vein (Owen, 1483
2011). The blood was refrigerated and allowed to coagulate for at least 24 hours prior to 1484
78
being centrifuged at 13000 RPM for 10 minutes. The resulting serum was removed for 1485
disease testing and the remaining material was frozen at -20°C for later genetic analysis. 1486
Additional samples were taken from boobook carcasses and shed feathers solicited 1487
from private citizens through BirdLife WA and a network of volunteers. Feathers were 1488
stored in paper envelopes at -20°C. All carcasses were stored frozen at -20°C until 1489
dissection when samples of muscle tissue were removed and stored in 100% ethanol for 1490
later analysis. 1491
Boobooks (n=137) were placed into one of six predefined regions based on 1492
similarities in geography and landscape type. The category ‘Exurbs’ (n=28) included 1493
individuals collected in areas immediately surrounding but not within the Perth 1494
Metropolitan area. ‘Perth Hills’ specimens (n=8) originated in an area of continuous forest 1495
east of Perth. Birds placed in the ‘Perth Metro’ category (n=71) originated in urban and 1496
suburban areas of Perth. Some boobooks were obtained from the Goldfields and Pilbara 1497
regions of Western Australia and were placed together in the ‘Remote WA’ (n=4) category. 1498
Boobooks from wetter, cooler, forested climates to the south of Perth were placed in the 1499
‘Southwest WA’ (n=17) category. The ‘Wheatbelt’ (n=9) category included all boobooks 1500
from highly-fragmented agricultural landscapes in the WA wheatbelt. 1501
1502
Genetic Analysis 1503
Microsatellites are commonly used in population genetic studies, particularly in bird 1504
species (Moura et al., 2017) for the purpose of individual fingerprinting, determining 1505
parentage, and exploring genetic variation and its spatial structure (Guichoux et al. 2011). 1506
Twenty microsatellite loci have been described for the Powerful Owl (Ninox strenua) and 19 1507
of these markers have been shown to be polymorphic in Australian Boobooks (Hogan et al. 1508
2009). Hogan et al. (2009) suggested these markers would be useful in genetic studies of all 1509
Ninox species tested. We used a subset of 15 loci microsatellites developed by Hogan et al. 1510
(2009) and after optimisation, nine (Nst02, Nst08, Nst11, Nst13, Nst14, Nst15, Nst16, Nst18, 1511
and Nst19; Table 4.1) were used to examine whether connectivity differed between the two 1512
types of anthropogenic landscapes. 1513
79
Table 4.1 The characteristics of the primers from 15 microsatellite loci amplified in Australian Boobooks (Ninox boobook) 1514 from Western Australia using primers adapted from (Hogan et al. 2007, 2009). 1515
For loci Nst02 and Nst18, PCR conditions were an initial denaturation step at 94 °C 1530
for 5 minutes, followed by 30 cycles of denaturation at 94 °C for 30 seconds, annealing for 1531
45 seconds at 50 °C, and extension of 45 seconds at 72 °C with a final extension of 5 minutes 1532
at 72 °C. This was followed by another 8 cycles of denaturation at 94 °C for 30 seconds, 1533
annealing at 53 °C for 45 seconds and extension at 72 °C for 45 seconds. The last cycle was 1534
followed by final extension at 72 °C for 10 minutes. All other loci PCR conditions were an 1535
initial denaturation step at 94 °C for 5 minutes, followed by 4 touch down cycles of 1536
denaturation at 94 °C for 30 seconds, annealing for 45 seconds at 60-54 °C, and extension of 1537
45 seconds at 72 °C with a final extension of 5 minutes at 72 °C. This was followed by 1538
another 25 cycles of denaturation at 94 °C for 30 seconds, annealing at 54 °C for 45 seconds 1539
and extension at 72 °C for 45 seconds. The last cycle was followed by final extension at 72 1540
°C for 5 minutes. This was followed by another 8 cycles of denaturation at 94 °C for 30 1541
seconds, annealing at 53 °C for 45 seconds and extension at 72 °C for 45 seconds. The last 1542
cycle was followed by final extension at 72 °C for 10 minutes. 1543
Statistical Analysis 1544
Data for boobook owls were analysed at nine microsatellite loci described by Hogan 1545
et al. (2007) and Hogan et al. (2009). However, one locus was excluded from analysis 1546
(Nst14) due to a high frequency of genotyping failures, leaving eight microsatellite loci 1547
available for use in the results presented here. Highly related individuals, known offspring 1548
(sensu Wang 2018), and boobooks of unknown geographic origin were removed from 1549
analysis resulting in a sample size of 137 adult individuals (Appendix 4.1). We conducted 1550
Mantel tests and spatial autocorrelation using a subset of these individuals. Boobooks of 1551
known regional origin lacking precise location data were removed. Additionally, when more 1552
than one individual was sampled at a single location (usually in the case of mated pairs) one 1553
individual was randomly removed from analysis. This left 124 individuals available for the 1554
Mantel test. To examine spatial autocorrelation a subset of these individuals from the Perth 1555
Metro, Exurbs, and Perth Hills regions (n=98) were used because of the high density of 1556
sampling within these regions. 1557
To examine genetic relationships among groups of individuals and potential 1558
populations we conducted analysis of molecular variance (AMOVA) and principal 1559
coordinates analysis (PCoA) in GenAlEx6.502 (Peakall and Smouse, 2012, 2006). GenAlEx 1560
81
was also used to calculate descriptive statistics for each region including mean number of 1561
alleles (NA), effective number of alleles (NE), mean observed heterozygosity across all alleles 1562
(HO), mean unbiased expected heterozygosity across all alleles (uHE), and fixation index (F). 1563
We also used GenAlEx to examine trends in isolation by distance using a Mantel test using 1564
all individuals with known coordinates and another Mantel test including only boobooks 1565
from the Perth Metropolitan area. GenAlEx was also used to calculate pairwise FST, pairwise 1566
Jost’s DST, and Nm between regions. Spatial autocorrelation was tested in GenAlEx using 1567
even sample classes of n=200. We assessed genetic structuring using the program 1568
STRUCTURE 2.3.4 (Hubisz et al., 2009) using a burn-in of 100,000 steps and a MCMC of 1569
1,000,000 steps. We conducted 20 runs each assuming a different number of genetic 1570
clusters (K=1-6). We used CLUMPAK (Kopelman et al., 2015) to visually depict STRUCTURE 1571
outputs. STRUCTURE HARVESTER Web v0.6.94 (Earl and vonHoldt, 2012) was used to 1572
estimate the most probable number of genetic clusters using the Evanno et al. (2005) delta 1573
K method. 1574
Results 1575
Direct Measurement of Dispersal 1576
Across all 12 records, the average recorded distance between original capture 1577
location and subsequent observation in fledgling and nestling boobooks was approximately 1578
10.5km with a maximum recorded movement of 52 km (Table 4.2). In our study, juvenile 1579
boobooks were observed moving an average of 8km and up to 26 km from their capture 1580
site. All the captures and re-sightings of nestlings and fledglings from our study occurred 1581
within the Perth Metropolitan area across urban and suburban habitat.1582
82
Table 4.2 Records of date a bird was tagged, its location, days and distances elapsed between capture and recovery of Australian Boobooks (Ninox boobook) banded as fledglings in Australia. 1583 Data from the Australian Capital Territory (ACT) and Queensland sourced from the Australian Bird and Bat Banding Scheme (http://www.environment.gov.au/science/bird-and-bat-banding). 1584 Western Australian data from re-sightings and recoveries of boobooks captured as part of this study. 1585
Date Banded State/Territory Days elapsed between capture
and recovery
Distance travelled (kms) between capture and
recovery
Recovery Method
29-November-1993 ACT 62 0 Found on highway/road; but not certainly hit by car
30-June-1994 Queensland 154 52 Band number read in field (bird not trapped)
04-December-1994 ACT 106 8 Collided with a moving road vehicle
13-January-2000 ACT 1709 18 Found dead, cause unknown
20-January-2001 ACT 78 4 Collided with a moving road vehicle
03-January-2004 ACT 165 4 Found dead, cause unknown
14-February-2008 ACT 21 0 Found sick or injured
10-November-2015 Western Australia 46 0 Found sick or injured
08-December-2015 Western Australia 986 12 Found sick or injured
11-December-2015 Western Australia 167 0 Band number read in field (bird not trapped)
31-December-2015 Western Australia 16 2 Found dead, cause unknown
17-January-2016 Western Australia 125 26 Band number read in field (bird not trapped)
1586
1587
Table 4.3 Analysis of Molecular Variance (AMOVA) results using six regional groups of Australian Boobooks (Ninox boobook) in Western Australia as populations. 1588
Source of variation Degrees of freedom
Sum of squares
Mean squares
Estimate of variance Variation (%)
Among Populations 5 37.794 7.559 0.048 1%
Within Populations 131 875.527 6.683 6.683 99%
Total 136 913.321
6.731 100%
83
1589
Figure 4.2 A corellogram showing genetic correlation values (r) as a function of distance (kms) using eight microsatellite 1590 markers in a subset of Australian Boobooks (Ninox boobook) n=98 from the Perth metropolitan area, adjacent exurban 1591 areas and the Perth Hills. U and L are 95% confidence intervals around the null hypothesis of no spatial genetic structure. 1592 No significant genetic structure is shown at any distance class. 1593
1594
1595
Figure 4.3 Principal coordinate analysis results based on eight microsatellite loci in Australian Boobooks (Ninox boobook) in 1596 Western Australia. Clustering does not correspond to potential populations and is driven by two common alleles and their 1597 heterozygotes at the locus Nst15. Blue = 161/161, Green = 161/uncommon allele, Purple = 163/161, Orange = 1598 163/uncommon allele, Red = 163/163, Black = no result. 1599
All results suggested that there was no meaningful spatial genetic structuring in the 1601
population of boobooks we sampled. The Mantel tests did not detect a meaningful 1602
correlation between genetic and geographic distances in the entire group of boobooks 1603
sampled (Rxy=0.046, p=0.194) or within the metropolitan area (Rxy= 0.082, p=0.070). This 1604
result was corroborated by spatial autocorrelation analysis of boobooks from the Perth 1605
Metro, Exurbs, and Perth Hills regions which did not indicate significant genetic structure at 1606
any distance class (Figure 4.2). PCoA initially showed three distinct genetic clusters with no 1607
apparent correlation with hypothetical regions, and the first two axes explaining only 1608
15.91% of variance (Figure 4.3). Interrogation of the data set revealed that the three 1609
clusters were defined by homozygotes of two common alleles and their heterozygotes 1610
(Figure 4.3). When the locus Nst15 was removed from the analysis, no clusters were 1611
discernible (Figure 4.4) and the first two axes explained only 14.01% of the variance. The 1612
apparent clusters when the locus Nst15 was included appeared to be a consequence of a 1613
combination of low allelic diversity at the locus Nst15 and little genetic structure in the 1614
other seven loci. A lack of genetic structuring was also indicated by AMOVA which 1615
determined that 99% of the total molecular variance was partitioned within regions and 1616
only 1% among regions (Table 4.3). Fixation index values for all regions were within or below 1617
the range reported in populations of another small owl species which were not found to be 1618
impacted by a genetic bottleneck (Proudfoot et al., 2006) (Table 4.4). Pairwise Fst values 1619
between regions were low overall with the highest values between the Remote WA region 1620
and the other regions, consistent with largest geographic distance (Table 4.5). Similar 1621
patterns were evident in the estimated number of migrants per generation between regions 1622
(Table 4.5). However, even the highest values detected were still relatively low, particularly 1623
when taking into context the substantial geographic distance between the Remote WA 1624
collection locations and other regions and the large geographic area over which Remote WA 1625
specimens were obtained. Pairwise Jost’s DST values were also low between regions with 1626
the only significant value detected between the “Exurbs” and “Perth Metro” regions (Dst = 1627
0.027, P= 0.015) (Table 4.6). The statistical significance of this value is likely to be an 1628
artefact of the substantially larger samples sizes of these regions rather than indicative of a 1629
meaningful biological difference in the alleles present in the two regions. STRUCTURE 1630
results did not show any spatial genetic clustering (Figure 4.5). Low Delta K values also 1631
85
support a lack of spatial genetic structure (Figure 4.6). A single genetic cluster was 1632
supported by mean LnProb values obtained using CLUMPAK (Appendix 4.2) and STRUCTURE 1633
HARVESTER (Appendix 4.3). 1634
1635
Figure 4.4 Principal coordinate analysis results based on seven microsatellite loci (i.e. no Nst15 – see Fig 3) in Australian 1636 Boobooks in Western Australia. No clustering is apparent across or within six sampled regions (“Exurbs” = areas 1637 immediately surrounding but not within the Perth Metropolitan area, “Perth Hills” = an area of continuous forest east of 1638 Perth, “Perth Metro” = urban and suburban areas of Perth, ‘Remote WA’ = Goldfields and Pilbara regions of Western 1639 Australia, “Southwest WA” = forested areas to the south of Perth, “Wheatbelt” = highly-fragmented agricultural landscapes 1640 existing primarily between the “Remote” region and all other regions). 1641
Co
ord
. 2
Coord. 1
Principal Coordinates (PCoA)
Exurbs
Perth Hills
Perth Metro
Remote WA
Southwest WA
Wheatbelt
86
1642
Figure 4.5 Visualization of Australian Boobooks (Ninox boobook) sampled from six regions in Western Australia (“Exurbs” = 1643 areas immediately surrounding but not within the Perth Metropolitan area, “Perth Hills” = an area of continuous forest 1644 east of Perth, “Perth Metro” = urban and suburban areas of Perth, ‘Remote WA’ = Goldfields and Pilbara regions of Western 1645 Australia, “Southwest WA” = forested areas to the south of Perth, “Wheatbelt” = highly-fragmented agricultural landscapes 1646 existing primarily between the “Remote” region and all other regions) using the STRUCTURE results from CLUMPAK 1647 comparing number of inferred genetic clusters (K) from 1-6. The data support a single genetic cluster. Each line represents 1648 an individual. The proportion of colours in each line represents the proportion of membership of each individual in each 1649 cluster. 1650
1651
87
1652
Figure 4.6 Plot of Evanno et al.’s (2005) delta K (ΔK) based on inferred genetic clusters (populations) ranging from 2 to 5 in 1653 Australian Boobooks (Ninox boobook) sampled from Western Australia. 1654
1655
Table 4.4 Genetic diversity parameters for Australian Boobooks (Ninox boobook) in six regions in Western Australia derived 1656 from eight microsatellite loci. Mean number of genotyped individuals (N), mean number of alleles per locus (NA), mean 1657 number of effective alleles (NE), mean observed heterozygosity (HO), mean unbiased expected heterozygosity (uHE). 1658
Table 4.5 Pairwise Fst and estimated number of migrants per generation (NM) between all geographic regions of Australian 1660 Boobooks (Ninox boobook) sampled in Western Australia. 1661
Region 1 Region 2 Fst Nm
Exurbs Perth Hills 0.015 16.7
Exurbs Perth Metro 0.012 19.8
Perth Hills Perth Metro 0.017 14.1
88
Exurbs Remote WA 0.037 6.5
Perth Hills Remote WA 0.054 4.4
Perth Metro Remote WA 0.039 6.1
Exurbs Southwest WA 0.016 15.8
Perth Hills Southwest WA 0.027 9.1
Perth Metro Southwest WA 0.012 20.2
Remote WA Southwest WA 0.033 7.3
Exurbs Wheatbelt 0.030 8.1
Perth Hills Wheatbelt 0.036 6.6
Perth Metro Wheatbelt 0.019 13.2
Remote WA Wheatbelt 0.043 5.5
Southwest WA Wheatbelt 0.019 13.0
1662
Table 4.6 Pairwise estimates of Jost's DST (below diagonal) and associated P values (above diagonal) for Australian 1663 Boobooks (Ninox boobook) sampled in five regions of Western Australia. 1664
Exurbs Perth Hills Perth Metro Southwest WA Wheatbelt
Exurbs
0.953 0.015 0.386 0.101
Perth Hills -0.049
0.679 0.562 0.485
Perth Metro 0.027 -0.016
0.235 0.363
Southwest WA 0.003 -0.011 0.011
0.752
Wheatbelt 0.041 -0.001 0.007 -0.028
1665
Discussion 1666
Both direct (banding) and indirect (genetic analysis) estimates of dispersal indicated 1667
widespread connectivity across all sampled populations despite extensive historical clearing 1668
of bushland in urban and agricultural landscapes. All statistical tests performed indicate a 1669
single admixed population of boobooks across all areas sampled. This result is consistent 1670
with a previous study which showed very little phylogenetic distinction between putative 1671
boobook subspecies across continental Australia (Gwee et al., 2017). The slightly higher Fst 1672
values observed between boobooks in the “Remote WA” group and other groups are likely a 1673
consequence of the group’s small sample size and the large geographic area from which the 1674
samples were derived. Alternately, weak isolation by distance across a large geographic 1675
area could explain this trend. 1676
The weak spatial genetic structuring both across Western Australia and within and 1677
between fragmented habitats is likely caused by effective movement between remnant 1678
habitat patches by dispersing juveniles. The genetic connectivity observed between 1679
89
fragmented landscapes and adjacent intact landscapes suggests historical movement 1680
between all areas despite extensive clearing over a long period of time (Saunders, 1989) 1681
while the observed capacity in our banding studies, of juvenile boobooks to disperse across 1682
substantial distances within fragmented urban landscapes, demonstrates that this type of 1683
habitat alteration does not constitute a barrier to juvenile dispersal. This result is consistent 1684
with dispersal patterns observed in other owl species. In a telemetry study of Burrowing 1685
Owls (Athene cunicularia), fledglings dispersed an average of 14.9 km (range 0.2 km - 53.1 1686
km) from their natal nest (Rosier et al., 2006). Similar dispersal patterns were observed in 1687
Spotted Owls (Strix occidentalis) (LaHaye et al., 2001). Within Australia, congeneric 1688
Powerful Owls (Ninox strenua) have been observed dispersing up to 18 km from their natal 1689
nest across “urban fringe habitat” (Hogan and Cooke, 2010). Dispersal by juvenile boobooks 1690
of distances substantially greater than those between patches of bushland habitat provides 1691
a plausible explanation for the lack of genetic structuring observed in the boobooks tested. 1692
While only movements within regions were observed in this study, the long distance 1693
contemporary dispersal observed within the Perth Metro region suggests the capacity for 1694
substantial post-breeding dispersal between regions. This result is consistent with the 1695
genetic estimate of migrants per generation among regions, suggesting considerable 1696
historical dispersal of juvenile boobooks (Table 4.5). 1697
Additionally, in the course of the study, boobooks were frequently observed and 1698
captured in urban areas outside of remnant bushlands. In some instances boobooks were 1699
observed successfully fledging young in areas where their home range would be expected to 1700
encompass no bushland whatsoever and be composed entirely of moderately dense 1701
suburban housing and light commercial development. If highly anthropogenically-altered 1702
habitats are able to support successful breeding attempts, these habitats likely constitute 1703
usable space despite their high degree of alteration and would not constitute a barrier to 1704
dispersal. Detection of moreporks (Ninox novaeseelandiae) at 80% of bushland patches in 1705
an urban area in New Zealand (Morgan and Styche, 2012) and documented use of highly 1706
developed suburban habitat by a female boobook during the non-breeding season (Olsen 1707
and Taylor, 2001) supports the hypothesis that these highly-altered habitat types do not 1708
provide a barrier to dispersal in boobooks. It is unclear to what degree the majority 1709
components of agricultural landscapes are “usable habitat” for boobooks but, on one 1710
90
occasion, in the course of this study, a boobook was observed hunting along a road >1km 1711
from any bushland, tree line, or patch of native vegetation, suggesting that boobooks 1712
actively utilize resources in habitats which we initially hypothesized to function as a hostile 1713
matrix between patches of usable habitat. In two Australian passerine species, natural 1714
history traits associated with tolerance of the “hostile matrix” in a fragmented landscape 1715
were demonstrated to correlate with spatial patterns of genetic diversity (Shanahan et al. 1716
2011). Boobooks are generalist predators capable of utilizing a wide variety of habitat types 1717
and are clearly capable of juvenile dispersal across urban development. Their capacity to 1718
use a wide variety of habitat types including highly anthropogenically-altered landscapes 1719
likely facilitates connectivity across ostensibly “fragmented” habitat. The lack of resistance 1720
observed in fragmented landscapes in our study of booboks probably protects them from 1721
the negative genetic impacts of fragmentation. Recent modelling of Mexican Spotted Owl 1722
(Strix occidentalis lucida) gene flow across fragmented habitats suggests that landscape 1723
resistance was an important predictor of genetic distance between populations for species 1724
with high dispersal capacity in highly fragmented landscapes (Wan et al., 2018). Owl species 1725
with more specialised habitat and dietary requirements including Blakiston’s Fish Owls 1726
(Bubo blakistoni) (Omote et al., 2015), Spotted Owls (Strix occidentalis) (Haig et al., 2001), 1727
and the more closely related Powerful Owl (Ninox strenua) (Hogan and Cooke, 2010) have 1728
shown genetic bottlenecks and potentially dangerous levels of inbreeding in response to 1729
habitat fragmentation. 1730
The lack of evidence for inbreeding or isolation as a consequence of habitat 1731
fragmentation does not necessarily imply that populations of boobooks in landscapes 1732
fragmented by urban and agricultural developments are demographically healthy or self-1733
sustaining. Weak spatial genetic structuring would likely also be observed in scenarios 1734
where fragmented habitats function as ecological sinks supported by healthy populations in 1735
adjacent intact habitats. This scenario is potentially even more likely in species with a high 1736
tolerance for altered habitats and substantial dispersal capacity. At least in urban areas, 1737
recent studies suggest that anthropogenic mortality from road strikes and secondary 1738
poisoning with anticoagulant rodenticides may pose significant threats to boobooks (Lohr, 1739
2018). Future work examining differences in life history parameters including adult and 1740
91
juvenile mortality across multiple habitat types would be useful in determining the relative 1741
utility of highly anthropogenically altered landscapes as boobook habitat. 1742
Genetic isolation and subsequent inbreeding could potentially become a problem for 1743
boobooks in urban and agricultural landscapes in the future despite their observed current 1744
dispersal capacity from banding studies if insufficient breeding hollows are retained at a 1745
landscape scale. Nest hollow availability is the key habitat requirement across the 1746
boobook’s range (Olsen and Taylor, 2001; Taylor and Canberra Ornitholgists Group, 1992) 1747
and urban fragments contain fewer hollow-bearing trees than intact forested areas (Harper 1748
et al. 2005). While nest hollow limitation does not currently appear to negatively impact 1749
boobooks in the Perth Metro area or WA wheatbelt (M. T. Lohr, unpublished data), 1750
continuing loss of nesting hollows through land clearing for additional development, 1751
inappropriate fire regimes, removal of nest trees for safety reasons, and urban infill could 1752
potentially reduce hollow availability in the future. In Powerful Owls (Ninox strenua), Hogan 1753
& Cooke (2010) detected instances of close inbreeding in two out of four pairs on the edge 1754
of urban areas near Melbourne despite a demonstrated capacity for dispersal up to 18km. 1755
Conversely, all three pairs nesting in continuous forested habitat were found to be 1756
unrelated (Hogan and Cooke, 2010). Hogan & Cooke (2010) speculated that this pattern 1757
could be explained by a lack of habitat for juveniles to disperse to, and subsequent 1758
clustering of related individuals, largely as a consequence of insufficient nest hollow 1759
availability. If patterns of boobook nest hollow availability ultimately approach those of 1760
Powerful Owls, this could lead to a reduction in genetic diversity and inbreeding depression 1761
over time in fragmented habitat types, even if boobooks are capable of dispersal between 1762
patches. However, if threatening processes and limiting factors in fragmented habitats are 1763
sufficiently addressed, both genetics and movement data suggest that boobooks should be 1764
capable of rapid recolonization and demographic recovery. 1765
Acknowledgments 1766
This project was supported financially by The Holsworth Wildlife Research 1767
Endowment via The Ecological Society of Australia, the BirdLife Australia Stuart Leslie Bird 1768
Research Award, and the Edith Cowan University School of Science Postgraduate Student 1769
Support Award. We thank Dr. Jamie Tedeschi for advice and technical assistance in 1770
laboratory work. We especially appreciate the contribution of boobook banding data by 1771
92
Jerry Olsen. Our research would not have been possible without contributions of samples 1772
and access to live birds provided by Kanyana Wildlife Rehabilitation, Native Animal Rescue, 1773
Native ARC, Nature Conservation Margaret River Region, Eagles Heritage Wildlife Centre, 1774
and many individual volunteers especially Simon Cherriman, Angela Febey, Amanda Payne, 1775
Stuart Payne, and Warren Goodwin.1776
93
Appendix 4.A A complete listing of the samples used in the analysis of microsatellite DNA polymorphisms, 1777
including the identification number (Individual ID), sample source, collection dates, collection locations 1778
(decimal lat/long), sampling locations/regions and age at sampling of Australian Boobooks used in this study. 1779
impact the abundance of tree hollows and may impact the entirety of the fragment 1840
depending on its size (Harper et al. 2005). In a survey of tree hollow occurrence in urban 1841
remnant woodlands in Melbourne, Australia, Harper et al. (2005) found no hollows in 12 of 1842
44 survey sites and 64% of remnants contained fewer than six hollow-bearing trees per 1843
hectare which is “well below that contained in areas of un-logged non-urban forest”. Urban 1844
remnant forests in Sydney were also found to have fewer hollow-bearing trees than 1845
continuous forest (Davis et al. 2014). The removal of large trees for timber and firewood as 1846
part of past management practices has also substantially decreased the number of hollow-1847
bearing trees in some urban remnants (Harper et al. 2005) and has directly impacted some 1848
102
threatened bird species such as the Swift Parrot (Lathamus discolor) (Webb et al., 2018). 1849
The exclusion of fire from urban forest fragments and removal of large trees due to safety 1850
concerns, may also play a role in reducing hollow formation and persistence (Harper et al. 1851
2005). Conversely, the inappropriate use of fire has been noted as a key driver of hollow 1852
loss (Stojanovic et al., 2016) and in agricultural regions has combined with other stressors 1853
such as intentional bulldozing of nest trees, and lone trees in paddocks being blown over as 1854
a consequence of greater exposure to wind (Saunders et al., 2014). 1855
Consistent long-term decline in abundance of large nest hollows used by endangered 1856
Carnaby’s Black-Cockatoos (Calyptorhynchus latirostris) has also been observed in remnant 1857
bushlands in agricultural landscapes in Western Australia (Saunders et al., 2014). The 1858
relative paucity of available hollows in landscapes which have been intensively altered by 1859
humans may be a limiting factor for wildlife which would otherwise be capable of using 1860
remnant bushlands and could be a factor contributing to overall declines in biodiversity. 1861
Nest Competition and Predation 1862
Even where nest hollow abundance is high, competition from introduced and 1863
overabundant native species can reduce nest hollow availability for obligate hollow-nesting 1864
wildlife. In North America, range-wide decline in three bluebird species (Sialia spp.) has 1865
been partially attributed to competition for nest hollows from European Starlings (Sturnus 1866
vulgaris) and House Sparrows (Passer domesticus) (Newton, 1994). In Australia, the 1867
introduction of hollow-nesting Common Mynas (Acridotheres tristis) was found to be 1868
correlated with declines in three native hollow-nesting bird species in the Canberra area 1869
(Grarock et al. 2012). Introduced European honeybees have been recorded as excluding a 1870
wide variety of native marsupial species from nest boxes in Australia (Beyer and Goldingay, 1871
2006). Galahs (Eolophus roseicapilla) and Western Corellas (Cacatua pastinator) are native 1872
to Western Australia but are overabundant in some areas and are believed to negatively 1873
impact endangered Carnaby’s Black-Cockatoos through competition for scarce nesting 1874
hollows (Johnstone et al., 2015; Saunders and Doley, 2017). Predation by the introduced 1875
Sugar Glider (Petaurus breviceps) in Tasmania, Australia is the key cause of the decline of 1876
the endangered hollow-nesting Swift parrot (Stojanovic et al., 2014). Understanding 1877
interactions between native and introduced hollow nesting species will be of increasing 1878
103
importance to conserving native biodiversity in areas where fragmentation simultaneously 1879
decreases hollow availability and facilitates growth of populations of introduced species. 1880
Impacts of Nest Boxes in Conservation 1881
As a response to hollow limitation, artificial nest hollow or “nest box” installation 1882
programs have been used for research aimed at understanding important life history traits 1883
of specific populations and have been used as an effective conservation measure to stabilize 1884
some declining populations (Lambrechts et al. 2012). These programs have been an 1885
important part of recovery efforts for hollow breeding birds worldwide. Routing of artificial 1886
hollows into living trees has been an integral part of successful efforts to increase 1887
abundance of the endangered Red-cockaded Woodpecker (Leuconotopicus borealis) in the 1888
southeastern United States (Walters, 1991). Widespread nest box provisioning efforts by 1889
private organizations have been widely attributed as a major factor in the recovery of three 1890
species of bluebirds in North America (Newton, 1994). Nest boxes have also been used to 1891
increase barn owl populations in Israel (Kan et al. 2013), Malaysia (Duckett and Karuppiah 1892
1990; Puan et al. 2012), and India (Parshad, 1999) as part of efforts to reduce crop damage 1893
by rodents. In Australia, construction of nest boxes is currently used successfully to mitigate 1894
losses of natural hollows for Carnaby’s Black-Cockatoos in the Western Australian 1895
agricultural zone (Johnstone et al. 2015) and Glossy Black-Cockatoos (Calyptorhynchus 1896
lathami) on Kangaroo Island (Mooney and Pedler, 2005) and has been used as a 1897
conservation tool in managing Critically Endangered Orange-bellied Parrots (Neophema 1898
chrysogaster) (Goldingay and Stevens, 2009) and Swift Parrots (Stojanovic et al., 2019). 1899
While most of these programs addressed lack of hollow availability, in some bird species, a 1900
variety of parameters impacting breeding success are higher in nest boxes than in natural 1901
hollows (Purcell et al. 1997). 1902
Nest boxes may not be a solution for all species, especially if nest hollow limitation is 1903
not the key cause of decline. For example, Loman (2006) found that nest hollow availability 1904
in small woodland patches was limiting for some obligate hollow-nesting passerine species 1905
but not others. In some instances, nest boxes may be preferred to natural nests and rapid 1906
adoption of nest boxes can give the appearance of nest limitation where there is none. For 1907
example, in one study, 83% of Tawny Owl pairs switched from natural nest sites to nest 1908
boxes within the year they were provided and 100% of pairs switched within four years but 1909
104
breeding density did not appear to change as a result of nest box provisioning (Petty, 1992). 1910
Purple Martins (Progne subis) in North America provide an even more extreme example of 1911
this dynamic. The eastern population of Purple Martins has used nesting structures 1912
provided by humans since prior to European colonization (Speck, 1941) and is now almost 1913
completely dependent on artificial nesting hollows constructed by humans (Morton et al. 1914
1990). While human-provisioned nest hollows clearly benefit this species, the potential 1915
risks of a population’s near-complete dependence on nest sites provided by humans are 1916
evident. In instances where nest boxes are preferred to natural hollows but are associated 1917
with lower nesting success they may even function as ecological traps (Klein et al. 2007, 1918
Heinshohn et al., 2015). Perhaps most fundamentally, nest box supplementation will not 1919
result in increases in abundance of the target species unless other resource requirements 1920
are already met (Durant et al. 2009). These factors should be considered before 1921
implementing or encouraging large-scale nest box programs and when evaluating the 1922
results of these programs. 1923
Knowledge Gaps 1924
Despite a large body of research on nest box impacts on native mammals and use of 1925
nest boxes in conservation efforts for cockatoos, few studies have focused on use of nest 1926
boxes by predatory birds in Australia. In a review of literature regarding nest box use by 1927
Australian bird species, only one of 17 species listed as having been studied was a predatory 1928
bird (Goldingay and Stevens, 2009). This study was conducted on a small hybrid population 1929
of boobooks on Norfolk Island and was an overview of conservation efforts rather than an 1930
empirical study of nest box impacts (Olsen, 1996). Another major knowledge gap relating to 1931
nest box impacts involves their use in developed areas. Less than 5% of Australian studies 1932
on the use of natural and artificial hollows have been conducted in urban landscapes 1933
(Durant, 2006). 1934
Despite the lack of studies relating to use of nest boxes by urban birds generally and 1935
Australian raptors specifically, artificial nest hollows have already been promoted as a 1936
conservation measure for urban raptors. Provision of nest boxes was suggested to improve 1937
Powerful Owl habitat in urban environments where scarcity of suitable nest hollows may be 1938
limiting abundance (Isaac et al. 2008). In one instance, subsequent localized nest box 1939
placement resulted in successful breeding of a nesting pair (McNabb and Keating, 2008; 1940
105
McNabb and Greenwood, 2011). While this particular effort was well justified and the 1941
result of this action is encouraging, the unregulated implementation of untested 1942
conservation actions intended to benefit sensitive species is concerning. Nest boxes 1943
intended for use by a wide variety of wildlife species are already commercially available 1944
from local businesses and instructions and plans are readily available online and are actively 1945
promoted for owls by the WA government Department of Biodiversity, Conservation and 1946
Attractions (Hussey, 1997). Both options are promoted as broadly beneficial to native 1947
wildlife despite a lack of rigorous testing for most species. Klein et al. (2007) suggested that 1948
correlation between increased breeding abundance and nest box provisioning should be 1949
proven prior to use of nest boxes as a conservation strategy. The widespread promotion 1950
and use of nest boxes necessitates studies addressing impacts of nest boxes on bird 1951
populations broadly and particularly on predatory birds and birds using urban areas. 1952
We studied the small owl, the Australian boobook in south-western Australia, as a 1953
model to examine whether nest box provisioning can increase occupancy by this species in 1954
human-altered landscapes. Australian boobook’s are an ideal study species as they are 1955
widespread but a 2015 report on population trends in Australian birds identified a serious 1956
decline in Australian boobook numbers from 1999-2013 and recommended that “further 1957
investigation is needed to understand the factors that are driving this consistent decline 1958
across regions” (BirdLife Australia, 2015). Nest hollow availability is believed to be the key 1959
habitat requirement across the boobook’s range (Olsen and Taylor, 2001; Taylor and 1960
Canberra Ornitholgists Group, 1992) and loss of tree hollows has been cited as one of the 1961
reasons for its decline in some areas (Debus, 2009). In the single published study involving 1962
nest box use by boobooks, lack of nesting hollows was implicated as one of the major 1963
factors contributing to the near extinction of Norfolk Island boobooks (Ninox 1964
novaseelandiae undulata) and nest boxes were a key tool used in its recovery program 1965
(Olsen, 1996). Boobook occurrence has been observed to correlate negatively with 1966
increased density of sealed roads and positively with forest cover, and nest hollow 1967
availability has been hypothesized as the factor driving differences in boobook abundance 1968
between urban and forested landscapes in and around Melbourne, Australia (Weaving et 1969
al., 2011). Urban fragments generally contain fewer hollow bearing trees than intact 1970
forested areas (Harper et al. 2005). Likewise, in the agricultural “wheatbelt” of Western 1971
106
Australia, loss of nest hollows is said to be one of the most important challenges facing 1972
wildlife conservation (Johnstone et al. 2015). Examination of patterns of nest box use by 1973
boobooks and its relationship with site occupancy across these two habitat types is 1974
necessary to understand their potential utility in the conservation of this species. 1975
Specifically we aimed to investigate whether Australian boobook occupancy in 1976
fragmented landscape types (agricultural and urban) was altered by providing nest boxes. 1977
Our hypothesis was that nest hollows would be limiting in agricultural and urban landscapes 1978
and that nest boxes would be quickly taken up by Australian boobooks. 1979
1980
Methods 1981
Study Sites 1982
To determine the impacts of nest box installation on site occupancy, surveys were 1983
conducted in 2015 at >30 sites each in each of three categories of land use: urban remnant 1984
bushlands, agricultural remnant bushlands, and areas of continuous bushland . Sites were 1985
located along the same approximate latitude in an area of south western Western Australia 1986
with a Mediterranean climate (Figure 5.1). Urban sites (n=35) found across the Perth 1987
metropolitan area were composed of bushland reserves managed by city governments or 1988
the Botanic Parks and Gardens Authority. Most sites were open woodlands dominated by 1989
Banksia sp., Eucalyptus gomphocephala, or E. rudis. Agricultural sites (n= 33) included both 1990
privately-owned bushlands and sites managed by the Western Australian Department of 1991
Biodiversity, Conservation and Attractions. All were within approximately 60km of the town 1992
of Kellerberrin, Western Australia. Dominant vegetation across these sites included Acacia 1993
acuminata, Eucalyptus capillosa, E. loxophleba, and E. salmonophloia. Continuous bushland 1994
sites (n=34) were located between the Perth Metropolitan area and areas of extensive 1995
agricultural development. They were bounded by the Great Eastern and Great Southern 1996
Highways to the North and the Brookton Highway to the South. Dominant vegetation in 1997
these sites was primarily Eucalyptus wandoo, E. marginata, and Corymbia calophylla. Intact 1998
bushland sites were included in surveys as a baseline against which to compare the efficacy 1999
of nest box supplementation as a management action intended to increase site occupancy. 2000
107
2001
Figure 5.1 Locations of survey sites in in southwestern Western Australia: urban landscapes in the Perth Metropolitan Area, 2002 continuous bushland in the Perth Hills, and agricultural landscapes within a 60km radius of Kellerberrin, Western Australia. 2003
Surveys 2004
In urban and agricultural bushlands, surveys were conducted 100m from a road or 2005
near the middle of the reserve in smaller reserves to reduce the impact of traffic noise on 2006
surveys. In continuous bushland areas, survey points were located approximately 5km apart 2007
to ensure independence. Baseline occupancy surveys were conducted in 2015 from 2008
September to December during the breeding season when boobooks call most frequently 2009
and are most easily detected (Olsen, 2011b). To maintain consistent detectability of 2010
boobooks, surveys were only conducted in the absence of rain and when estimations of 2011
wind speed were below a score of 3 on the Beaufort scale. Surveys consisted of passively 2012
listening for boobook vocalizations from a fixed point for 15 minutes followed by five 2013
minutes of intermittent broadcast of recorded boobook vocalizations in accordance with 2014
methodology used by Liddelow et al. (2002) to survey nocturnal birds in south-western WA. 2015
Immediately following the survey, the area was scanned using a 1000 lumen LED headlamp 2016
to detect any boobooks that had been attracted by the calls but had not vocalised. All sites 2017
108
were classified as “occupied” or “not occupied.” A subsequent round of surveys was 2018
conducted at all sites in September-December of 2016 after the installation of nest boxes to 2019
determine occupancy using the same methodology used during the previous breeding 2020
season. 2021
Nest box construction and placement 2022
Nest boxes were constructed using recycled, 18mm form-ply, a waterproof and long-2023
lasting material used mainly for concrete construction work. Each box consisted of a 2024
wooden cube measuring 300mm long and 300mm wide, and a depth ranging from 450mm 2025
at the front to 500mm at the rear, creating a forward-sloping roof. The dimensions of the 2026
box were chosen to reflect dimensions of active boobook nests observed by the authors and 2027
to deter Galahs, which prefer deeper boxes and are aggressive competitors for nest hollows. 2028
A hollow log-round of diameter 120-200mm was attached to the front of each box using 2029
screws fastened from the inside. This served to create a ‘verandah’ designed to protect the 2030
internal nest-chamber from weather, and also to prevent non-target species with 'heavy-2031
chewing' behaviour (e.g. Galahs) from enlarging the box entrance hole and potentially 2032
destroying it. A wooden lid with a c. 50mm overhang was attached with a hinge fitted to the 2033
rear, and the sides were reinforced with aluminium flashing, again to prevent chewing 2034
species from destroying the lid. Boxes were assembled in such a way to leave ’air slots’ 2035
~15mm wide beneath the lid on both sides, designed to facilitate air-flow and subsequent 2036
internal temperature fluctuation to deter feral honey bees, which have specific hive 2037
temperature requirements of 32-35˚C, from taking up residence. Two coats of pale-green, 2038
water-based exterior paint were applied to all external surfaces, to protect the sawn 2039
wooden edges from the elements and thus defer deterioration, and to help the boxes blend 2040
in with the natural environment. A layer of coarse woodchips c. 150mm deep was added to 2041
the inside of each box to create an internal nest chamber consisting of well-drained 2042
substrate that allows hollow-nesting species to scrape a shallow bowl in which eggs are 2043
deposited. Wood chips consisted of c. 20mm diameter pieces and were collected near 2044
installation sites. 2045
Nest boxes were installed in trees with multi-strand, galvanised wire (‘clothesline 2046
wire’) c. 4mm thick, threaded through plastic/rubber pipe (‘hosepipe’) to protect the tree's 2047
bark from wire damage. The installation process was carried out using the following steps: 2048
109
1) a small loop was created at one end of the wire, and the tail end threaded into one of two 2049
c. 8mm holes pre-drilled on the rear surface of the box, at each of its top corners; 2) the tail 2050
end was then threaded out through the second hole and the wire pulled through until it was 2051
tight; 3) after being threaded through a length of hosepipe, the main length of wire was 2052
looped horizontally around a solid, vertical section of trunk, being passed above an oblique 2053
or horizontal limb used to ‘hang’ the box and prevent it sliding down (Figure 5.2); the tail 2054
end was then threaded through the small loop at the back of the box and twitched into 2055
place for secure attachment. Sufficient length of wire was used so each box was ’strung’ 2056
firmly but not hung in such a way that left wire tightly constricting on the trunk. This 2057
method is similar to the ‘habisure system’ described in Franks and Franks (2006), and it 2058
ensures secondary (horizontal) growth of the tree’s trunk (i.e. limb thickening) can take 2059
place naturally. Permanent attachment methods involving fixings such as coach bolts or 2060
screws were avoided to 1) minimise injury to the tree’s vascular cambium that may lead to 2061
unnecessary infection or damage, and 2) ensure boxes were not ‘pushed off' as the tree 2062
trunk expands during secondary growth, resulting in the potential collapse of an occupied 2063
nest-site and/or a safety risk to passers-by. 2064
110
2065
Figure 5.2 Attachment system used to hang nest boxes used in this study. 2066
Nest boxes were placed in fifteen sites in both urban and agricultural remnant 2067
bushlands which did not have boobook detections in the previous round of surveys. Nest 2068
boxes were installed in February 2016 shortly after the termination of the breeding season 2069
to allow adequate time for detection by boobooks prior to the following breeding season. 2070
All nest boxes were placed in the nearest suitable tree to the survey point in all 30 2071
experimental sites. All nest boxes were hung at a height below 11m to facilitate observation 2072
of their contents and greater than 4m because most published records indicate minimum 2073
Figure 5.3 A nest box installed in one of the remnant bushlands in an agricultural landscape in Western Australia. 2076
Nest Box Monitoring 2077
We examined the contents of all nest boxes for evidence of use by boobooks or 2078
potentially competing species. Nest box contents were viewed using a video camera 2079
(MiGear ExtremeX Sports Action Camera) mounted on an 8m telescoping fiberglass pole to 2080
record video footage of the inside of each nest box. All videos were viewed at the nest site 2081
112
to ensure that adequate footage of the nest boxes’ interior was obtained to allow 2082
identification of contents. Videos were retained for later review. Nest boxes were checked 2083
on three occasions during the breeding season in 2016 (July 24-26, October 7-9, and 2084
November 18-25) and once during the 2017 breeding season (September 27-29). In the 2085
2017 surveys, a single nest box at one of the urban sites was unavailable to be checked as it 2086
had been destroyed by a bushfire. 2087
Statistical Analysis 2088
We compared differences in territory occupancy in 2015 across all three habitat 2089
types prior to treatment using a Chi-square test with a post hoc pairwise test of 2090
independence for nominal data. We used McNemar's Chi-squared tests with continuity 2091
correction to examine differences in occupancy between years at treated and untreated 2092
sites. All tests were performed using RStudio 1.1.383 (RStudio, Inc., Boston, MA, USA). 2093
Results 2094
Prior to nest box treatment, boobooks were more commonly detected in continuous 2095
bushland sites (85.3%, n = 34) than in remnant bushlands in urban (30.8% n = 39) and 2096
agricultural (21.2%, n = 33) landscapes. Occupancy rates were significantly greater at 2097
continuous bushland sites than urban sites (p<0.001) or wheatbelt sites (p<0.001) but did 2098
not differ between urban and wheatbelt sites (p=0.517). No significant differences in 2099
occupancy were detected between years in any of the treated or untreated groups across all 2100
three habitat types (Table 5.1). However, non-significant increases in occupancy occurred in 2101
sites provided with nest boxes in both fragmented habitats while non-significant declines 2102
occurred in control sites in both urban and agricultural habitats (Table 5.1).2103
113
2104
Table 5.1 Annual change in occupancy of Australian Boobooks at continuous bushland sites and sites with and without supplemental nest boxes in remnant woodland in urban and agricultural 2105 landscapes in Western Australia. 2106
Urban/Periurban 14/90 15.6 Injury status Wild 4/42 9.5 0.580
In care/dead 13/88 14.8
Season Summer 3/60 5.0 *0.024
Autumn 8/35 22.9
Winter 3/12 25.0
Spring 2/22 9.1 AR exposure 0-0.01 mg/kg 3/23 13.0 0.759
0.01-0.10 mg/kg 2/8 25.0
0.10-0.50 mg/kg 5/22 22.7 >0.50 mg/kg 3/12 25.0
2554
130
2555
Figure 6.1 Seasonal Toxoplasma gondii seroprevalence in Australian Boobooks (Ninox boobook) in Western Australia. Width 2556 of the bars is representative of sample size. 2557
2558
131
2559
Figure 6.2 Toxoplasma gondii seroprevalence in meat juice from deceased Australian Boobooks (Ninox boobook) in Western 2560 Australia in four different categories of anticoagulant rodenticide exposure (A= ≤ 0.01 mg/kg, B=0.01 mg/kg – 0.10 mg/kg, 2561 C 0.10 mg/kg - 0.50mg/kg, D ≥ 0.50mg/kg) Width of the bars is representative of sample size. 2562
Discussion 2563
Apparent seroconversion in the two individuals which tested negative in serum 2564
samples but positive in meat juice samples may be an artefact of the MAT test used. False 2565
negative results can be obtained during acute stages of T. gondii infection because the test 2566
is only sensitive to IgG antibodies, and not IgM antibodies which are present at the onset of 2567
infection (Sroka et al., 2008). It is entirely plausible that the two boobooks were 2568
experiencing active infections when initially sampled but their infections would only be 2569
detected by the subsequent meat juice sampling. We believe that this explanation, in 2570
combination with the lack of a significant difference in seropositivity rates between serum 2571
and meat juice samples justifies our decision to combine data from both matrix types in the 2572
other analyses. 2573
132
Studies using the same commercially available modified agglutination test on other 2574
continents have found varying rates of seropositivity across multiple raptor species: 34.5% 2575
in the south-eastern USA (n=281) (Love et al., 2016), 25.7% in Taiwan (n=206) (Chen et al., 2576
2015), 35.8% in France (n=53) (Aubert et al., 2008), 50.0% in Portugal (Lopes et al., 2011), 2577
and 10.7% in Italy (n=93) (Gazzonis et al., 2018). While at the lower end of the scale, our 2578
results (13.1%) were within the ranges previously reported in studies of predatory birds. 2579
Interestingly, overall seropositivity was nearly identical to the rate of 13.0% reported for a 2580
native marsupial carnivore (chuditch, Dasyurus geoffroii), in Julimar Valley (Parameswaran, 2581
2008), an area of continuous bushland adjacent to our sites in the Perth Hills. 2582
Four potential explanations exist for the relatively low seropositivity rates we 2583
observed. Australian boobooks have not previously been evaluated using this test and it is 2584
possible that species-specific factors may have led to false negative results. We view this 2585
scenario as unlikely because, while false negatives using MAT are common in some species – 2586
particularly in dogs (Liu et al., 2015) – this test has been used successfully to detect 2587
toxoplasma seropositivity in a wide variety of other predatory bird species (Chen et al., 2588
2015; Gazzonis et al., 2018; Lopes et al., 2011). 2589
It is also unlikely that our use of meat juice in addition to serum would have reduced 2590
detections relative to other studies. A study directly examining detection of T. gondii 2591
antibodies in meat juice did not find reduced detectability or degradation of antibodies in 2592
response to repeated freezing and thawing of meat (Mecca et al., 2011). If anything, the 2593
use of this methodology should have increased seropositivity detection in our study relative 2594
to other studies which tested only serum. The numerically but not significantly higher 2595
detection rate of T. gondii antibodies in meat juice samples is consistent with this 2596
hypothesis. 2597
The diet and trophic position of boobooks may also provide some explanation for 2598
the relatively low seropositivity rates seen in boobooks. Australian Boobooks are medium-2599
sized owls (Olsen, 2011a) and consume a variety of invertebrate and vertebrate prey (Trost 2600
et al., 2008). A study on T. gondii seropositivity in wild birds in Spain found seropositivity 2601
rates ranging from 0% to 25% in six small and medium sized owl species (Cabezón et al., 2602
2011). However, the same study detected T. gondii antibodies in 68% of all Eurasian Eagle-2603
133
owls (Bubo bubo) (Cabezón et al., 2011). This species is substantially larger and occupies a 2604
higher trophic level than the other owl species tested. In the context of this research, the 2605
seropositivity of boobooks we observed is typical of an owl species of its size and diet. 2606
Landscape Type 2607
The relatively warm and dry climate in our study areas may explain our observation 2608
of lower seropositivity rates than in most other areas of the world where raptors have been 2609
sampled. Worldwide, toxoplasma prevalence is lowest in hot arid areas, presumably due to 2610
shorter duration of oocyst viability under hot dry conditions (Meerburg and Kijlstra, 2009). 2611
In a study examining habitat impacts on T. gondii seropositivity in wild rabbits, seropositivity 2612
was substantially higher in habitats with more shade and humidity (Almería et al., 2004). 2613
This pattern may explain the lack of significant difference observed in seropositivity 2614
between landscape types. Counter-intuitively, clearing of land for urban and agricultural 2615
uses could lead to a reduction in T. gondii seroprevalence despite potential increases in cat 2616
abundance if the reduction in vegetative cover results in an increase in soil temperature and 2617
decrease in soil moisture, leading to inhibition of T. gondii oocyst viability. Future work 2618
examining T. gondii seroprevalence in a single intermediate host species across paired 2619
habitat types in areas with substantially different rainfall levels would be useful in 2620
determining relative contributions of cat abundance and soil moisture to seropositivity in 2621
intermediate hosts. 2622
Age 2623
We were surprised that no difference in seropositivity was detected between age 2624
classes. Some studies have found that T. gondii detection increases with age in wild 2625
predatory birds (Cabezón et al., 2011; Lindsay et al., 1993; Lopes et al., 2011) which is in 2626
keeping with the hypothesized lifelong persistence of the parasite after infection. However, 2627
other studies of predatory birds (Gazzonis et al., 2018) and wild rabbits (Oryctolagus 2628
cuniculus) (Almería et al., 2004) have failed to detect a difference in seropositivity between 2629
different age classes but did not address why no correlation was detected. It is possible that 2630
our grouping of boobooks into two coarse age classes of < one year and ≥ one year obscured 2631
longer-term trends in seropositivity. Because boobooks are relatively long lived – one was 2632
re-sighted in the field alive nearly 16 years after it was originally banded (Commonwealth of 2633
134
Australia, 2015) – they are potentially at risk from the long-term effects of latent T. gondii 2634
infection similar to those observed in humans. Impaired reaction time resulting from T. 2635
gondii infection is cumulative in humans and increases with the duration of latent infection 2636
(Havlícek et al. 2001). Similar effects in long-lived wildlife could predispose individuals to 2637
greater risk of predation, accident, or vehicular collision. Use of predatory bird species with 2638
a greater number of more easily-identifiable age categories (such as Wedge-tailed Eagles 2639
(Aquila audax)) could help to resolve questions relating to both changes in seropositivity 2640
between age classes and whether cumulative impacts of latent toxoplasmosis are 2641
problematic for predatory birds. 2642
Injury Status 2643
The lack of significant difference in seropositivity between boobooks found dead or 2644
held by wildlife carers and those captured in the wild was also unexpected and runs 2645
contrary to observations by Hollings et al. (2013) in Tasmanian pademelons (Thylogale 2646
billardierii) shot for pest control purposes and pademelons killed by motor vehicle collisions. 2647
It is unlikely that this is a consequence of our testing of multiple matrix types, as 2648
seropositivity was numerically – though not significantly – higher in meat juice samples from 2649
deceased boobooks. It seems more likely that our inclusion of boobooks which were killed 2650
or disabled by a wide variety of causes may have obscured any potential effect specific to 2651
motor vehicle collisions. In humans, reduced concentration time and increased reaction 2652
time were proposed as the mechanisms by which T. gondii seropositivity increased rates of 2653
car accident (Flegr et al., 2002). It is unlikely that these potential causative factors are 2654
relevant to all the causes of mortality or injury associated with the boobooks in our study. 2655
Unfortunately, uncertainty over proximate causes of death or injury in the boobooks we 2656
tested precluded direct testing of a more specific relationship with seropositivity. 2657
Season 2658
Several potentially interacting biological factors could plausibly explain the higher 2659
seropositivity rates of boobook samples obtained in autumn and winter. Boobooks 2660
consume a higher proportion of mammals and birds in winter relative to other times of year 2661
when insects make up a larger percentage of their diet (Trost et al., 2008). Additionally, 2662
temperature and rainfall patterns in autumn and winter in southwest Western Australia are 2663
more conducive to T. gondii oocyst viability and, as a consequence, infection rates in prey 2664
135
species may be higher at this time of year. A similar explanation was given for observations 2665
of higher T. gondii seroconversion rates observed in house cats in autumn and winter 2666
relative to spring and summer (Simon et al., 2018). Both factors may increase the risk of 2667
boobooks consuming prey with tissue cysts containing T. gondii bradyzoites and subsequent 2668
infection in winter. Boobooks with recent infections may also have been easier to capture 2669
or more likely to die and be injured and consequently be sampled by our study, increasing 2670
the detected seropositivity of individuals sampled at this time of year. 2671
Anticoagulant Rodenticide Exposure 2672
Alternately, exposure to anticoagulant rodenticides may have contributed to 2673
increased seropositivity of samples obtained in winter. Significantly higher liver 2674
concentrations of anticoagulant rodenticides have been observed in boobooks in the Perth 2675
metropolitan area in winter relative to spring and summer (Lohr, 2018). Additionally, while 2676
the Fisher’s exact test failed to detect a significant difference between rodenticide exposure 2677
categories, the seroprevalence of boobooks in the lowest exposure category with 2678
insubstantial amounts of rodenticide was numerically lower than the three categories with 2679
clinically relevant AR exposure (≥0.01 mg/kg) (Figure 6.2). Sub-lethal exposure to 2680
anticoagulant rodenticides has been found to correlate with immune dysfunction in bobcats 2681
(Lynx rufus) (Serieys et al., 2018) and has been hypothesized as the explanation for an 2682
observed correlation between anticoagulant rodenticides and notoedric mange (Riley et al., 2683
2007). These correlations are to some degree called into question by a study on domestic 2684
cats (Felis catus) which did not find a substantial link between anticoagulant rodenticides 2685
and immune dysfunction (Kopanke et al., 2018). However, even if immunosuppression is 2686
not the mechanism by which AR exposure facilitates hyper-parasitism, similar increases in 2687
pathogen and parasite load correlated with AR exposure have also been documented in 2688
Great Bustards (Otis tarda) exposed to the AR chlorophacinone (Lemus et al., 2011). 2689
If AR exposure facilitated reactivation of latent toxoplasmosis, this could explain the 2690
increase in seroprevalence detected in winter and autumn. Alternately, it is possible that a 2691
synergistic interaction between AR exposure and T. gondii infection increased the 2692
probability of the boobooks dying and entering this study to be tested. A synergistic effect 2693
on probability of mortality involving the AR chlorophacinone and the pathogen Francisella 2694
tularensis has been suggested in common voles (Microtus arvalis) (Vidal et al., 2009). 2695
136
Another hypothesis potentially explaining the possible correlation between AR exposure 2696
and seropositivity is that boobooks could simply be exposed to both T. gondii and ARs at 2697
higher rates in winter months as a consequence of rodents making up a higher proportion of 2698
their diet at this time of year. These hypotheses are not mutually exclusive and would be 2699
difficult to distinguish without experimental study of this dynamic in a laboratory setting. 2700
While boobooks showed few significant trends in T. gondii seropositivity, this may be 2701
primarily an issue of low statistical power to detect such trends caused by relatively low 2702
sample sizes. This is a common problem when studying cryptic, nocturnal, carnivores which 2703
occurr at low densities and are difficult to capture. However, seasonal differences in 2704
seropositivity suggest that conditions influencing oocyst viability may be a more important 2705
determinant of exposure risk than the factors we examined directly. Future work evaluating 2706
the utility of boobooks and other raptors as bioindicators of environmental T. gondii 2707
contamination should examine seropositivity rates across temperature and rainfall 2708
gradients. The use of boobooks as bioindicators could help identify important landscape-2709
level drivers of T. gondii prevalence and has the potential to inform management actions 2710
and translocation efforts intended to benefit susceptible native mammals. 2711
Acknowledgments 2712
This project was supported financially by The Holsworth Wildlife Research Endowment 2713
via The Ecological Society of Australia, the BirdLife Australia Stuart Leslie Bird Research 2714
Award, and the Edith Cowan University School of Science Postgraduate Student Support 2715
Award. We thank Annette Koenders, Adriana Botero, and Louise Pallant for advice and 2716
technical assistance in serology testing. Our research would not have been possible without 2717
contributions of samples and access to live birds provided by Kanyana Wildlife 2718
Rehabilitation, Native Animal Rescue, Native ARC, Nature Conservation Margaret River 2719
Region, Eagles Heritage Wildlife Centre, and many individual volunteers especially Simon 2720
Cherriman, Angela Febey, Amanda Payne, Stuart Payne, and Warren Goodwin. 2721
2722
137
Chapter 7 Summary, Synthesis, and Management Implications 2723
I examined four distinct potential threatening processes which I predicted had the 2724
potential to vary in magnitude of impact between habitats fragmented by urban and 2725
agricultural land uses. Australian Boobooks (Ninox boobok) did not appear to be 2726
substantially negatively impacted by lack of nest hollow availability, infection with 2727
Toxoplasma gondii, or genetic isolation in either landscape type. However, I did detect 2728
considerable exposure to anticoagulant rodenticides (ARs) associated with the use of 2729
habitats containing commercial and residential development. 2730
In this chapter, I highlight the most important findings of each chapter and 2731
contextualise their relevance to their respective fields outside of the specific system I 2732
studied. I also discuss the contribution of my research to the theoretical framework in 2733
which the impacts of habitat fragmentation are typically evaluated. I then suggest specific 2734
practical implications of my findings for management actions intended to maintain or 2735
increase native biodiversity in landscapes dominated by intensive human land uses. 2736
Summary of major findings: 2737
Objective 1. Critically review literature on anticoagulant rodenticide exposure in native 2738
wildlife in Australia to clarify its role as a threatening process. 2739
My review of the literature relating to anticoagulant rodenticides in Australia 2740
revealed widespread anecdotal accounts of both primary and secondary anticoagulant 2741
rodenticide (AR) poisoning among a taxonomically diverse group of non-target wildlife. Key 2742
differences between Australia and other developed nations were noted in the regulation of 2743
ARs. Most notably, second generation anticoagulant rodenticides (SGARs) are readily 2744
available for purchase without a license in Australia, unlike in the United States and Canada. 2745
Australia is also one of only two countries to allow the use of the first generation 2746
anticoagulant rodenticide (FGAR), pindone and to allow its use in widespread repeated 2747
baiting of natural systems for control, rather than eradication, of introduced species. 2748
Additional research is recommended to evaluate this practice. The review also identified 2749
patterns in world literature relating to reptiles and rodenticides which suggest the potential 2750
for high tolerance to rodenticides in at least some reptile taxa. Further experimental testing 2751
is necessary to determine if this hypothesized resistance makes reptiles efficacious vectors 2752
138
of ARs to humans and predatory wildlife. If so, rodenticide poisoning in warmer areas of the 2753
world with diverse and abundant reptile herpetofaunas, may be a greater threat to 2754
predatory wildlife than in the cool temperate regions where most AR ecotoxicology work 2755
has been conducted. 2756
Objective 2. Investigate the relationship between exposure to anticoagulant rodenticides 2757
and urban and agricultural fragmentation. 2758
Exposure to anticoagulant rodenticides (ARs) was prevalent in the boobooks tested 2759
(72.6%) and higher than typically observed in similar studies of predatory birds on other 2760
continents. The vast majority of the rodenticides detected were the more persistent second 2761
generation anticoagulant rodenticides (SGARs). AR exposure correlated positively with 2762
proximity to urban/periurban habitat at all spatial scales and negatively with use of 2763
agricultural areas and native bushland. The association between AR exposure and the 2764
proximity of boobooks to urban and suburban development (but not agricultural land uses), 2765
supports modelling which suggests that matrix type can exert strong influences on wildlife 2766
inside habitat patches (Sisk et al., 1997). The strongest correlations between AR exposure 2767
and habitat were found at the spatial scale of a boobook’s estimated home range. This 2768
suggests that predatory birds with larger home ranges may be at risk of AR exposure over a 2769
larger proportion of the landscape. Additional research on non-target AR exposure in 2770
Australia is urgently needed to determine the level of threat posed to other wildlife species, 2771
particularly carnivores and scavengers with large home ranges which are already listed as 2772
threatened (e. g. quolls (Dasyurus sp.) and Tasmanian devils (Sarcophilus harrisiii). 2773
Objective 3. Determine if urban and agricultural fragmentation influence boobook genetic 2774
structure. 2775
Boobooks did not exhibit substantial genetic structure among landscapes dominated 2776
by urban development, agricultural crops, or native bushland in between. This trend held 2777
with the inclusion of boobook samples originating across a larger geographic area including 2778
the majority of Western Australia. Banding data from my study and others demonstrated 2779
that fledgling boobooks are capable of dispersing across urban habitats for distances far 2780
greater than those between remaining bushland fragments. In combination, these findings 2781
suggest a high degree of landscape permeability and genetic connectivity in boobooks 2782
across all areas sampled. Highly mobile species have a greater probability of survival than 2783
139
less mobile species in areas which have experienced habitat fragmentation (Ewers and 2784
Didham, 2006). High mobility despite fragmentation coupled with the apparent capacity to 2785
use matrix habitat in at least some circumstances likely explains the persistence of 2786
boobooks in highly fragmented landscapes, albeit at lower densities. 2787
Objective 4. Examine whether nest box supplementation increases site occupancy at 2788
unoccupied sites and whether this effect differs between urban and agricultural landscapes. 2789
Boobooks occupied fewer sites in urban and agricultural remnant bushlands than in 2790
continuous woodland. Nest box supplementation at unoccupied sites did not alter site 2791
occupancy over the duration of this study. However, one nest box in an urban bushland 2792
remnant was successfully used by a boobook. Nest hollows do not appear to be a limiting 2793
factor in the use of remnant woodlands by boobooks in either fragmented landscape type 2794
despite boobooks being obligate hollow nesters. Nest box supplementation is unlikely to be 2795
an effective tool for increasing boobook abundance in remnant woodlands but anecdotal 2796
observations of boobooks utilising nest boxes in urban areas completely devoid of native 2797
bushland suggest that nest boxes may reduce matrix hostility and increase usable space in 2798
highly-altered areas lacking remaining suitable tree hollows. 2799
Objective 5. Explore patterns of Toxoplasma gondii seropositivity in boobooks across the 2800
urban, agricultural, and natural landscapes. 2801
Toxoplasma gondii seropositivity did not vary significantly among urban, agricultural, 2802
and woodland dominated landscape types. Most other factors which other studies have 2803
found to correlate with T. gondii seropositivity (i.e. age, season, injury status, and exposure 2804
to environmental pollutants) did not show significant correlations. Failure to detect these 2805
trends may have been caused by insufficient statistical power associated with low 2806
seropositivity rates. However, higher seropositivity was observed in cooler wetter seasons. 2807
This trend could be related to environmental conditions favouring oocyst viability, greater 2808
Wolf, A., Cowen, D., Paige, B., 1939. Human toxoplasmosis: occurrence in infants as an 4082
encephalomyelitis verification by transmission to animals. Science (80-. ). 89, 226–227. 4083
https://doi.org/10.1126/science.89.2306.226 4084
Young, J., De Lai, L., 1997. Population declines of predatory birds coincident with the 4085
introduction of Klerat rodenticide in north Queensland. Aust. Bird Watch. 17, 160–167. 4086
Yu, L., Shen, J., Su, C., Sundermann, C. a., 2013. Genetic characterization of Toxoplasma 4087
gondii in wildlife from Alabama, USA. Parasitol. Res. 112, 1333–1336. 4088
https://doi.org/10.1007/s00436-012-3187-0 4089
4090
4091
185
Co-author Statements 4092
4093
Signed co-author statements verifying my role in the production of papers and manuscripts which 4094
make up chapters in this thesis are provided in this section. 4095
Chapter 2 4096
4097
186
Chapter 4 4098
4099
187
Chapter 5 4100
4101
188
Chapter 6 4102
4103
189
Copies of original publications 4104
I include below copies of the first page of published peer-reviewed journal articles corresponding to 4105 chapters in this thesis. No licenses are required to reproduce these papers either in part or in full 4106 when included as part of a PhD thesis per the Elsevier license agreement: 4107 https://service.elsevier.com/app/answers/detail/a_id/565/track/AvMKOAoHDv8W~QaHGnwa~yKg_4108 38qZS75Mv9z~zj~PP_6/ 4109