The Decline, Fall, and Rise of a Large Urban Colonising Bird

Context The process of urbanisation results in dramatic landscape changes with long-lasting and sometimes irreversible consequences for the biota. Urban sensitive species can be eliminated from the landscape, while urban tolerant species can persist in or colonise the changed environment. Objectives Here we used historical atlas data to examine the changing distribution of the Australian Brush-turkey, a recent urban colonising species, at continental and city scales, and the changing land use in urban areas occupied by the species. Methods We assessed changes at the continental scale from 1839-2019. We then assessed colonisation of the cities of Sydney and Brisbane, located 900 km apart, over the period 1960-2019. At the city scale, we quantied the changing land use within Brush-turkey occupied areas over time using classication of satellite imagery.


Introduction
Urbanisation is a major land use change with often dramatic and long-lasting consequences for biodiversity (McDonald, Kareiva, & Forman, 2008;McKinney, 2002). Natural habitat is largely cleared and replaced by extensive areas of built structures, roads, and other impervious surfaces, while remaining vegetation is often highly fragmented (Grimm et al., 2008). Urban dwelling animals must also contend with increased levels of chemical and sensory pollutants, altered trophic interactions, potential competition with or predation from non-native species, and exposure to anthropogenic sources of 2020; Jones & Göth, 2008), as has been observed for other generalists (Croci, Butet, & Clergeau, 2008;Moller, 2009). Resolving the contrast between their apparent success in urban areas and possible decline in rural areas requires a greater understanding of how the Brush-turkey distribution has changed over time in different regions.
In this study we investigated temporal shifts in Brush-turkey distributions at multiple scales. To investigate the Brush-turkey distribution at the continental scale, we mapped Brush-turkey sightings records and quanti ed changes in occupied bioregions for six time periods from 1839 to 2019. To investigate the Brush-turkey distribution at the city scale, we quanti ed the number of Brush-turkey occupied suburbs for four time periods from 1960 to 2019 for the cities of Sydney and Brisbane. We further investigated if changes to urban land-use at the suburb scale corresponded with changes to the urban Brush-turkey distribution.

Study Area
The study area for the continental-scale assessment includes approximately 3000km tract of the Australian east coast, from the tip of Cape York, Queensland, to the Southern Highlands, New South Wales, and up to 750km inland. The city-scale assessment focuses on the cities of Sydney and Brisbane, which are the largest cities within their respective states of New South Wales (NSW) and Queensland as well as the two largest cities within the estimated range of the study species. Sydney has a population of 5.37 million people (ABS, 2021) and a temperate climate (Herron et al., 2018). Brisbane has a population of 2.56 million people (ABS, 2021), and a subtropical climate. Both cities are bordered by national parks to the north, west, and southeast. Both cities contain many smaller patches of remnant vegetation alongside managed arti cial greenspaces (Keith, 2004;Lunney, Hutchings, & Hochuli, 2010;Museum, 2003)

Historic and Recent Sightings
We downloaded all Brush-turkey occurrence records from the Atlas of Living Australia, as well a taxonspeci c citizen science project, Big City Birds (formerly BrushTurkeys), on 25/3/2021. The Atlas of Living Australia (hereafter "Atlas") is a collaborative digital platform that compiles Australian taxon occurrence records from multiple sources, including government databases, citizen science records, and museum collections, and provides information on data quality. Big City Birds is a targeted citizen science project that collected detailed ecological data on Brush-turkeys, including presence, counts, sex, and behavioural observations such as nesting and roosting locations (Hall, Martin, Burns, & Hochuli, 2021). These records were downloaded separately as they are not currently included in the Atlas.
Records with no latitude and longitude coordinates and records with no exact date were excluded. We included records prior to 1900 if the sighting had an exact year. We further eliminated records that were not human observations of a wild bird, nest, chicks/juveniles, or eggs, or museum specimens with a collection location. Records indicating captive animals, such as those in zoos or pets, were removed. Data from one source, Queensland WildNet (hereafter "WildNet"), was redownloaded directly from the source website on 26/7/21 (Queensland Government, 2021) due to inaccuracies identi ed in the dates assigned to WildNet records in the Atlas.
To further lter the data, we mapped all remaining occurrence records using ArcMap 10.8. We spatially ltered data by removing records over water and records where the provided location description did not match GPS coordinates. Outliers were visually identi ed and eliminated if they met all of the following rules: the record was from an opportunistic survey by non-experts (e.g. citizen science data), the record was outside of previously published distribution maps for the species (Birdlife International, 2021; (Göth et al., 2006), and there was no other sighting within the same bioregion or within 500km for the given time period. Four remaining outliers were also eliminated: a 14/7/2000 record from Diamantina National Park was removed as it was considered a misidenti cation by Ley, Tynan, and Cameron (2011); two Questagame records from Adelaide, South Australia, in 2019 were removed due to a lack of any other sightings from a highly populated urban area indicating these sightings are likely spurious; and one 1/12/1949 record from Adelaide, was eliminated as this record was of preserved eggs only with an uncertain collection location.

Continental Scale Distribution Time Series
To investigate distribution changes at the continental scale, we assessed the number and location of Brush-turkey records across bioregions. Bioregions represent large-scale environmental divisions of Australia based on common climate, soil, geology, and vegetation (DAWE, 2020; Thackway & Cresswell, 1995). Bioregions are commonly used as coarse landscape units in management and conservation at the regional scale. We categorised records temporally into six time periods : pre-1900, 1900-1939, 1940-1959, 1960-1979, 1980-1999, and 2000-2019. The longer time intervals for the periods prior to 1939 re ect the scarcity of records for the species compared to the more recent time periods.
To estimate the changing Brush-turkey range at the continental scale for each period, we plotted all ltered Brush-turkey records in ArcMap10.8 and aggregated sightings points into a single polygon. We used an aggregation distance of 1000km for the three earlier time periods and 500km for the more recent time periods to re ect the increasing availability and coverage of sightings data, resulting in a more conservative estimation of the Brush-turkey range in the more recent periods. We used the smooth polygon tool with Berzier interpolation algorithm to eliminate sharp angles and added a 0.1 decimal degree (11.1km) buffer to ensure sighting points were contained within the interior resulting polygon. We refer to the resulting polygon as the "estimated range".
To visualise how the number and density of records changed in each bioregion over time, the estimated range polygon was then intersected with the bioregions layer and spatially joined with the Brush-turkey records point layer. We display the total number of records for each bioregion and time-period.

City Scale Distribution Time Series
Page 6/22 To assess Brush-turkey distribution changes in the major urban centres of Sydney and Brisbane, we de ned the spatial extent of each city using the "signi cant urban area" (SUA) classi cation used by the Australian Bureau of Statistics (ABS). SUAs represent contiguous large urban centres or clusters of related urban centres (ABS, 2017). The Sydney and Brisbane SUAs were subdivided into suburbs, representing o cially recognised and named localities within cities and towns (ABS, 2019). To map changes in Brush-turkey suburb occupancy, we plotted all ltered sightings records for the two cities in ArcMap 10.8 for each time-period. We intersected the city suburbs layer with the sightings point layer to produce a density map displaying a count of Brush-turkey records in each suburb for each time-period.
To investigate change at ner temporal scales, we further categorised records from 1960-2019 into 5-year intervals and measured the number of Brush-turkey occupied suburbs for each city. Records prior to 1960 were not included due to a scarcity of records for both cities.

Land Use Analysis
To assess changing patterns of land use within Brush-turkey occupied areas in Sydney and Brisbane, we downloaded Landsat satellite imagery of the two cities from 1979, 1999, and 2019. These years were chosen as they represent the end-dates of the 1960-1979, 1980-1999, and 2000-2019 time periods used for the Brush-turkey range analysis. 1979 imagery was obtained from the Landsat 2 multispectral scanner, 1999 imagery from the Landsat 5 thematic mapper, and 2019 imagery from the Landsat 8 operational land image and thermal infrared scanner. Images were selected from July through September to obtain comparable images with minimal cloud cover. The satellite images were imported into ArcMap 10.8 and clipped to shape les of the Sydney and Brisbane SUAs.
We used supervised image classi cation with a maximum likelihood approach to quantify land use in each city for each year (see Fischer, Edwards, Weber, Garnett, & Whiteside, 2021; Hahs & McDonnell, 2006). Twenty training samples were manually assigned for each of the following land use classes: commercial, residential, dense vegetation, open greenspace, and bare land. We then ran the supervised image classi cation tool in ArcMap 10.8 to classify all images. The majority lter and boundary clean tools were used to remove isolated pixels and ragged boundaries, producing a more generalised output map. The classi ed images were then intersected with the Brush-turkey occupied suburbs layer to quantify the amount of each land use type within the Brush-turkey range for each city, in each year. Chisquared tests of association were conducted in IBM SPSS 26 to test for associations between year and land use within Brush-turkey occupied suburbs. Post hoc Z-tests with Bonferroni corrections were conducted to determine which land use classes had signi cantly different proportions between time periods.

Occurrence Records
A total of 116433 Brush-turkey occurrence records were collected from 34 different sources. Following data ltering, this was reduced to 98019 for the period 1839-2019 (Table 1). Of these, 69671 records were located within signi cant urban areas. The three largest contributing sources of Brush-turkey records were eBird (53.2%), the O ce of Environment and Heritage Atlas of NSW Wildlife (24.6%), Big City Birds (8.3%).  (Fig. 1). Three additional bioregions were continuously occupied from 1900 to 2019: Central Mackay Coast, Nandewar, and New England Tablelands (Fig. 1b-f).
These bioregions are mostly situated along the east coast of Australia, with a few located inland in Central NSW and Southern Queensland.

Continental Scale Distribution
Historical records prior to 1900 (n=68) report Brush-turkey presence in eleven bioregions, covering a total area of 1018348km 2 . In two of these, the inland Cobar Peneplain and NSW Southwestern Slopes, Brush-turkey records disappeared after 1900 and were not observed again in any period (Fig. 1a). Occupied bioregions increased to 12 during the 1900-1939 period, with additional records in the Central Mackay Coast, Mitchell Grass Downs, Nandewar, and New England Tablelands. However, records ceased in the NSW Southwestern Slopes, Southeastern Highlands, and Cobar Peneplain bioregions (Fig. 1b). This represented an overall decrease in area of the total Brush-turkey range to 1005970km 2 . Ten bioregions were occupied during the 1940-1959 period, with Brush-turkey records disappearing from the inland Queensland bioregions of Einasleigh Uplands and Mitchell Grass Downs (Fig. 1c). This represented a decrease in total area of the Brush-turkey range to 684500km 2 . Overall, from 1900-1959, the estimated Brush-turkey range expanded in the central-western part of the species distribution but receded from the north-western and south-western part of their pre-1900 distribution. The total area of the estimated Brushturkey range declined by 333848km 2 during this period, or a decrease in total area of 32.8%, from the pre-1900 baseline.
Brush-turkey were recorded within 13 bioregions during 1960-1979, with new records from the Desert Uplands and Kanmantoo bioregions, and records resuming in the Einasleigh Uplands (Fig. 1d). This represents a total increase in area of the Brush-turkey range to 956441km 2 . During the 1980-1999 period, the number of occupied bioregions increased to 16, with new records from the Mulga Lands and Gulf Plains. Records also resume in the South-eastern Highlands which had not had any records since before 1900 (Fig. 1e). This represented a further increase in total Brush-turkey range area to 1138037km 2 . Brushturkey occupied bioregions decreased to 14 in 2000-2019, with records disappearing from the Mulga Lands and Desert Uplands (Fig. 1f). This represented a decrease in the total Brush-turkey range to 1124483km 2 .
Overall, from 1960-2019, the estimated Brush-turkey range appeared to have contracted in the centralwestern part of their distribution but has expanded in the north-west and south-east. From 1960-2019, the total area of the estimated Brush-turkey range increased by 168042km 2 , or 17.6%, and by 106135km 2 , or 10.4%, from the pre-1900 baseline.

City Scale Distribution
In both Sydney and Brisbane, Brush-turkey occupied suburbs show a large overall increase during the period 1960-2019 (Fig. 2), with the largest increase in occupied suburbs occurring in the last decade. In Sydney, the number of occupied suburbs remained consistently low, uctuating between 1-3 suburbs, across the 5-year time intervals from 1965 to 1994, this increased rapidly through to the present day (2019), to a total of 310 suburbs. In Brisbane, occupied suburbs increased steadily in each 5-year interval from 1965 to 1989 to a total of 50, then decreased from 1990-1994 to 28. The number of occupied suburbs then increased through all 5-year intervals till 2019, to a total of 289 suburbs (Fig. 2).
In Brisbane, Brush-turkey occupied suburbs during the 1960-1979 period were scattered across the central-western part of the city, with more isolated records in the north and southeast. The highest percentage of sightings came from suburbs in the centre and central west of the city (Fig. 3a). During the 1980-1999 period an increasing number of suburbs were occupied in the central, southwestern, and northern parts of the city, as well as the suburbs of the southeast (Fig. 3b). By 2000-2019, Brush-turkeys occupied the majority of suburbs in northern and central Brisbane, with a further increase in the number of occupied suburbs in the south. The suburbs with the largest percentage of records were in the centralwest part of the city, while large areas in the south of the city remained unoccupied (Fig. 3c).
In Sydney, Brush-turkey records rst appeared in the north and northwest of the city during the 1960-1979 period (Fig. 3d). Increasing numbers of northern suburbs were occupied during the 1980-1999 period, as well as suburbs in the west of the city, however Brush-turkeys disappeared from the north-western suburbs (Fig. 3e).

Discussion
After experiencing a catastrophic human-mediated decline in the early 20th century, particularly in and around urban areas, Brush-turkeys are actively recolonising the large global cities of Brisbane and Sydney. Their urban expansion was not limited to existing greenspaces, as we found that Brush-turkeys have increasingly colonised less vegetated and more developed suburbs over recent decades, indicating that the species is now thriving in urban areas. The changes in the Brush-turkey distribution were not limited to urban areas, with this species spreading consistently into rural and natural areas since 1960. Firstly, our assessment identi ed an estimated decrease in occupied land area of about 33% from 1900-1959, followed by an overall estimated increase of land area of about 18% from 1960-2019. Most of the land area lost was in the western and southwestern portions of the species' range, while expansion occurred in the northwest.

Continental Distribution Trends
Brush-turkeys expanded into one new bioregion, the Gulf Plains, in the last few decades. The Gulf Plains bioregion lies along the north coast of Australia and is adjacent to the Cape York Peninsula and Einasleigh Uplands bioregions, which have been continuously and near continuously occupied by Brushturkeys respectively. Given the steady increase in records over the last four decades, it is likely that Brushturkeys naturally expanded their range into the Gulf Plains bioregion from source populations in the adjacent bioregions as these neighbouring populations grew over recent decades.
There is a general trend of increasing Brush-turkey records from 1900-2019 in all continuously occupied bioregions, with the greatest increase in the number of observations occurring in the 2000-2019 period (Table S2). However, the increase in observations was more common in coastal and urban bioregions with larger human populations. The largest increase in records was in the NSW North Coast, South Eastern Queensland, Sydney Basin, and Wet Tropics bioregions, where the total number of records increased by several orders of magnitude. While Brush-turkey records increased over time in the Brigalow Belt South, Nandewar, and New England Tablelands bioregions, the number of records in these regions did not increase at a comparable rate to the four previously mentioned bioregions. This may be due to a lower human population in these areas, and hence fewer sightings, or it may indicate a smaller Brushturkey population.
Brush-turkeys ceased to be reported from multiple bioregions, indicating the long-term local extinction of this species from those areas. Three of these bioregions, Cobar Peneplain, South Western Slopes (NSW), and Riverina are at the southwestern edge of the species' distribution. A fourth bioregion, Desert Uplands, is to the northwest of their range. Previous research suggested that Brush-turkey numbers were in decline in these areas (Göth et al., 2006), including the threatened population in the Nandewar and Brigalow Belt Estimation of the Brush-turkey range, and associated changes to the population, was considerably more di cult for earlier time periods due to the scarcity of records. Older records were primarily based on opportunistic reporting, incidental observations, museum specimens, and records collected from various published and unpublished literature. The rst concerted national survey of Australian birds was conducted between 1977-1981 for the rst Atlas of Australian Birds (Blakers, Reilly, & Davies, 1984). It is likely that our data, based on records prior to this survey, underestimated the Brush-turkey range for these time periods. Additionally, some uncertainty exists for records prior to 1960. While all records used in this study included an exact date and coordinates, this was often based on the central point of a grid or locality for many records prior to the introduction of GPS technology (1980's). The continental scale used in this study may also obscure population changes at a local scale. Despite these limitations, the presence and absence of records at the scale of entire bioregions can be clearly observed with the data used in this study. We can thus be con dent of the broad scale changes in the Brush-turkey distribution.

Urban Distribution Trends
Prior to the 1960s, Brush-turkey records were conspicuously absent from Sydney and Brisbane, despite the existence of records from other parts of the Sydney Basin and South Eastern Queensland bioregions. Anecdotal evidence suggests that the species did occur in these areas prior to European colonisation and prior to the 20th century (Jones & Göth, 2008) and this aligns with the species existence in neighbouring bioregions. This indicates that Brush-turkeys became locally extinct around these urban centres prior to their return across both cities from the mid-1960s.
The rapid increase in Brush-turkey observations in the inner suburbs of Brisbane and Sydney in the 2000-2019 period counters an earlier suggestion that cities may act as ecological traps for the species due to reduced reproductive success in urban areas (Jones & Everding, 1991). The total number of suburbs occupied in both cities has greatly increased over the last 20 years, continuing previously identi ed trends (Göth et al., 2006;Jones et al., 2004), and our results show that Brush-turkey occupied suburbs have become less vegetated and more built-up over time: more urban. Part of this change is due to changing land use within previously occupied suburbs, however the majority of this change has been driven by a rapid, more than tenfold, increase in the total suburban area occupied by the species across both cities. More recently colonised suburbs have been overall less vegetated and more developed than earlier colonised suburbs. This trend is likely to continue as Brush-turkeys colonise new suburbs that are more remote from source populations in rural and natural areas at the city edge.
Emigration from populations in natural areas, translocations, and recruitment from local nesting have all been suggested as mechanisms for dispersal into suburban areas (Jones & Everding, 1991). Our results support the idea that new arrivals from non-urban populations are likely to be a source of replenishment for Brush-turkey populations in suburbs adjacent to bushland (Jones et al., 2004). Older occupied suburbs are geographically closer to non-urban Brush-turkey records in both Brisbane and Sydney.
However, the importance of colonisation from non-urban populations likely diminishes in inner-city suburbs that are more distant from natural woodland areas. In these areas, the dense urban matrix constrains dispersal making movement from distant areas more di cult (Canedoli, Manenti, & Padoa-Schioppa, 2018;Fischer & Lindenmayer, 2007). Brush-turkeys have been reported to have a high juvenile mortality rate (Göth & Vogel, 2002), however their continued spread across Brisbane and Sydney indicates that urban breeding success is not only maintaining the population but su cient to support the expansion of the population. While we anticipate that Brush-turkeys will continue to spread into currently unoccupied urban suburbs and increase in density, it remains to be seen if they will spread from the cities into unoccupied neighbouring vegetation. However, the establishment of a thriving Brush-turkey population after introduction on Kangaroo Island (Jones and Göth, 2008) demonstrates the versatility and adaptability of the species in novel environments, making this a strong possibility.
On the surface, Brush-turkeys do not t the pro le of the typical successful urban dwelling bird. Their ground nesting, lack of parental care, and poor ight abilities contrast with the off-ground nesting and high dispersal ability (through ight) common to most successful urban birds (see Bressler et al., 2020;Evans, Chamberlain, Hatchwell, Gregory, & Gaston, 2011;Moller, 2009). However, they have a generalist omnivorous diet (Jones & Göth, 2008), and have been observed feeding on novel food resources including introduced plants and anthropogenic food sources (Brookes, 1919;D. N. Jones & Everding, 1991). Brush-turkeys are also considered to be a highly disturbance tolerant species (Blumstein, 2006), and show reduced fear behaviour in urban areas compared to reserves and natural bushland (Hall et al., 2020). A generalist diet and increased boldness are common traits among urban birds Moller, 2008) and likely help the species colonise and persist in urban areas where disturbances and unnatural food sources are common. This suggests that a species may only need a few urban suitable traits to thrive in urban areas under the correct conditions; this highlights that a wider list of candidate species may be able to effectively colonise cities.
The increasing movement of Brush-turkeys into less vegetated suburbs will likely lead to more frequent encounters with suburban residents. Proximity of wildlife and humans can occasionally lead to situations of human-wildlife con ict, particularly in urban areas (Soulsbury & White, 2015). Brush-turkey presence in suburban areas has resulted in complaints from residents in response to damage to gardens caused by foraging and the construction of 3 tonne nest mounds, both of which involve raking soil and leaf litter (Jones & Everding, 1991). Complaints regarding chasing pets and small children, stealing food, noise, and fouling have also been reported (Jones & Göth, 2008), leading to calls for the management of predominantly urban Brush-turkey populations. While Brush-turkeys are not considered to be a threatened species across their distribution, their apparent decline in the western and southwestern ends of their range necessitates careful consideration when managing their population in urban areas. Further surveys of the connectivity between urban and non-urban Brush-turkey populations are needed to determine the importance of urban habitats as refugia for the species.

Limitations of Atlas and Citizen Science Data
The dramatic increase in the number of Brush-turkey records in the new millennium is facilitated by the creation of citizen science projects and community uptake of smartphone apps and online platforms for reporting wildlife observations ( into the Brush-turkey population, including ner scale presence and habitat preferences, will be possible over the next 20-years with increasing citizen science participation. A particular challenge of observing trends using citizen science data is disentangling a biological signal from the effects of human population and the accompanying spatial and temporal clustering of observations (Isaac et al., 2014). The greatest increase in the number of Brush-turkey reports was in the Sydney Basin and Southeast Queensland bioregions. These areas also have the highest human population densities within the Brush-turkey range (ABS, 2021). Importantly, irrespective of the accessibility of reporting sightings, it is clear from the data that Brush-turkeys have spread across urban Brisbane and Sydney over recent decades. However, the upwards trend in the number of observations appears to begin in some bioregions as early as the 1960s, where modern technologies to assist citizen scientists did not exist. Given the increasing availability of methods and technology to record and report observations, the complete absence of Brush-turkey records from some bioregions in the 2000-2019 period is strong evidence that the species is rare or absent from these areas, including the regions where they were historically observed.

Conclusions
Understanding how species distributions have changed over time and in response to human derived land use change is foundational to inform their management and conservation, as well as predicting how they will respond to future landscape changes. Using a blend of historic records, ecological surveys, and citizen science data our study determined that the Brush-turkey has undergone a complex range shift over the past century, disappearing from the edges of their range in the southwest while recolonising the heavily modi ed urban areas on the Australian east coast. Over the last sixty years, the species has successfully colonised more built-up and less vegetated areas of Brisbane and Sydney. The Brush-turkey has become an incredibly successful urban dwelling species, despite its specialised reproductive strategy and poor dispersal ability, broadening scienti c understanding of the traits that can characterise successful city dwelling wildlife. Future research should focus the following areas: tracking ner scale distribution changes for the species at continental and city scales, particularly at the edges of their range; determining the drivers behind Brush-turkey declines in the western parts of their range and expansion in urban areas; and quantifying Brush-turkey responses to different land uses within urban areas.