The Bird Assemblage of the Darwin Region (Australia): What Is the Effect of Twenty Years of Increasing Urbanisation?

: There has been considerable urban development in the Darwin region over the last twenty years; as for most fauna in Australia since colonisation, the potential effects to the bird assemblage were expected to be disastrous. To provide a broad overview of changes, bird survey data from 1998 and 2018 were extracted from BirdLife Australia’s ‘Atlas of Australian Birds’ database. A total of 165 species were categorised into primary food source feeding guilds and levels of food specialisation. This was integrated into ArcGIS along with land use change mapping from 1998 and 2018 to investigate its impact on bird assemblages. There was no signiﬁcant change in overall species numbers when all sites were analysed. However, when sites were separated into those with increased urbanisation or decreased greenspace, several sites showed a signiﬁcant change in the number of species. For the majority of species, analysis of primary food types found no difference in the proportion of species within the assemblages between 1998 and 2018, regardless of the level of urbanisation or greenspace; the exception being those species that primarily feed on insects, where the difference was just signiﬁcant. An analysis using bird community data sorted into levels of food specialisation also found no difference between 1998 and 2018 despite habitat changes. These ﬁndings suggest that although there has been considerable urban development in the Darwin region, bird communities are remaining relatively stable.


Introduction
Global population trends show that humans are moving from rural areas into cities at a considerable rate, and once sparsely populated regions are being transformed to cope with the influx of people [1,2]. Subsequently, urbanisation is now widely considered a major threat to biodiversity conservation [3][4][5].
Species' responses to urbanisation can vary significantly [6]; however, prior to the 1990s, very few ecological studies were undertaken in urban areas due to them being considered unviable habitats for faunal populations and, ergo, immaterial from a conservation standpoint [7]. It is now shown that urban areas support an array of species that have been able to tolerate or adapt to the highly fragmented new environments but, due to a lack of appropriate knowledge, the success of urban conservation programmes may be hampered [4]. There is little doubt that avian assemblages are greatly altered by urbanisation; many bird species unable to adapt to the suddenly changed environment will often move out of an area resulting in diminished biodiversity and allowing invasive species to move in [4,8,9]. Other species can benefit from urbanisation; these are often Diversity 2021, 13,294 2 of 18 birds with a more generalist diet and life-history traits that are conducive to living in a fragmented habitat [10].
Whilst studies of bird populations in urban areas abound in the temperate zones of both the Northern and Southern hemispheres, research into tropical urban avian communities is scant [11]. This trend is continued in Australian studies with the majority of research being undertaken in the temperate regions, predominantly in the two heavily populated states of New South Wales and Victoria. In contrast, this study is situated in the monsoonal tropics of northern Australia; specifically, in Darwin, the capital of the Northern Territory (NT). There has been targeted research on a range of species and thorough overviews of bird distributions within the region [12][13][14], with more detailed studies of shorebirds [15][16][17][18] and mangrove assemblages [19][20][21]. To date, however, there have been no studies of trends in terrestrial bird assemblages as the city has grown. Given the increased rate of urbanisation in the global tropics [22], coupled with high levels of biodiversity in the zone [11,[23][24][25], there is an increasing need to gather data to better understand how bird communities are coping with the rising encroachment of human habitation. This is particularly interesting in Darwin as, unlike all other urban centres in Australia, Darwin has no introduced birds, so all the adaptation is being undertaken by native species as their environment is changed.
In the nearly 45 years since Darwin was severely damaged by Cyclone Tracy, where 70-80% of dwellings were destroyed [26,27], the city has grown from a perceived 'frontier town' to be a modern capital city. This development has seen an increase of over 20 new suburbs in the greater Darwin area and surrounds, and land once considered bush or rural properties is now being subdivided into urban blocks. One of the most common effects of urbanisation is the increasing prevalence of exotic species [28][29][30]; however, this has not occurred in the Darwin region with only four species listed as 'foreign invaders', none of which have established permanent breeding populations [31].
In this paper, we shall test this theory by investigating the response of the terrestrial avian assemblages in the Darwin region to urban expansion and land use change over the past 20 years.

Study Area
Darwin (12.4634 • S, 130.8456 • E) is situated on the north coast of Australia, a landscape dominated by tropical savanna (Figure 1).
The population of approximately 140,000 [32] constitutes an increase of nearly 60,000 people in twenty years [33]. The climate is monsoonal and experiences distinct annual dry (May to September) and wet (November to April) seasons with transitional periods in between. Mean annual rainfall is approximately 1700 mm; the mean minimum and maximum temperatures range from 19.3 • to 25.3 • and 30.6 • to 33.3 • , respectively [34]. Compared to other major Australian capital cities, housing density is low; under 20 private dwellings per square kilometre as opposed to between 150 to 200 per square kilometre in Sydney and Melbourne [35]. A combination of urban and periurban environments in the Darwin region provides resources for avian populations that are typically unavailable outside of this area in the dry season. The population of approximately 140,000 [32] constitutes an increase of nearly 60,000 people in twenty years [33]. The climate is monsoonal and experiences distinct annual dry (May to September) and wet (November to April) seasons with transitional periods in between. Mean annual rainfall is approximately 1700 mm; the mean minimum and maximum temperatures range from 19.3° to 25.3° and 30.6° to 33.3°, respectively [34]. Compared to other major Australian capital cities, housing density is low; under 20 private dwellings per square kilometre as opposed to between 150 to 200 per square kilometre in Sydney and Melbourne [35]. A combination of urban and periurban environments in the Darwin region provides resources for avian populations that are typically unavailable outside of this area in the dry season.

Spatial Data
To investigate broad-scale changes in urbanisation, Landsat satellite imagery of the study area from April 1998 and 2018 was obtained from the United States Geological Survey's Global Visualization Viewer (GloVis), with 1998 imagery obtained from the Landsat 4 and 5 Thematic Mapper (TM) satellite and 2018 imagery taken by the Landsat 8 Operational Land Imager (OLI) and Thermal Infrared Sensor (TIRS) ( Table A1 and Table A2). April marks the end of the wet season in Darwin, allowing for images with minimal cloud cover and maximum vegetation growth. Images were clipped in ArcGIS version 10.4.1 [36] to a shapefile of Darwin region localities provided by the Northern Territory Government Department of Environment and Natural Resources and then imported into SAGA GIS version 7.3.0 [37]. Classification of land use types was carried out via geographic object-based image analysis (GEOBIA). In traditional pixel-based image classification, classes are assigned per pixel; however, GEOBIA uses segmentation and classification to better replicate what the human eye perceives [38,39]. The segmentation process combines pixels of similar spectral properties into objects in the form of polygons, and these can

Spatial Data
To investigate broad-scale changes in urbanisation, Landsat satellite imagery of the study area from April 1998 and 2018 was obtained from the United States Geological Survey's Global Visualization Viewer (GloVis), with 1998 imagery obtained from the Landsat 4 and 5 Thematic Mapper (TM) satellite and 2018 imagery taken by the Landsat 8 Operational Land Imager (OLI) and Thermal Infrared Sensor (TIRS) (Tables A1 and A2). April marks the end of the wet season in Darwin, allowing for images with minimal cloud cover and maximum vegetation growth. Images were clipped in ArcGIS version 10.4.1 [36] to a shapefile of Darwin region localities provided by the Northern Territory Government Department of Environment and Natural Resources and then imported into SAGA GIS version 7.3.0 [37]. Classification of land use types was carried out via geographic objectbased image analysis (GEOBIA). In traditional pixel-based image classification, classes are assigned per pixel; however, GEOBIA uses segmentation and classification to better replicate what the human eye perceives [38,39]. The segmentation process combines pixels of similar spectral properties into objects in the form of polygons, and these can then be classified using either supervised or unsupervised techniques [40]. After automated segmentation in SAGA GIS, the two clipped images were classified by assigning 'training sites' (essentially selecting a minimum number of polygons and ascribing them a land use type) and then running a supervised classification. The resulting vector layers were then manually edited using the original satellite image to reassign any misclassified polygons. The initial uncorrected GEOBIA and user-corrected images were then re-imported back into ArcGIS, where fifty accuracy assessment points were randomly generated and an error matrix was constructed to assess both the producer (SAGA) and user (human) accuracy when assigning classification. For all map classifications satellite imagery, aerial photographs and Google Earth Pro version 7.3.2.5776 (64 bit) were used to assist in the Diversity 2021, 13, 294 4 of 18 accuracy assessment; however, due to the retrospective nature of the earlier imagery, only the 2018 images could be further checked, if required, using ground control points.

Bird Survey Data
Survey data were extracted from the BirdLife Australia 'Atlas of Australian Birds' database (hereafter referred to as the 'Bird Atlas') for the years 1998 and 2018. Several types of surveys compiled these data: systematic bird surveys of 2 ha, 5 km and 500 m; unstandardised bird surveys, either along a fixed route or incidental; bird list surveys and the Shorebird 2020 surveys (a record of shorebird sightings not necessarily in coastal habitats). All records include a location, latitude and longitude, dates and species common names. In most records, a time is recorded and whether there is any breeding activity. Sighting notes of interest are sometimes included.
As the focus of the project was on terrestrial, predominantly diurnal species, Bird Atlas records were excluded if the species was almost exclusively nocturnal, was a waterbird or seabird (except Magpie Geese, Anseranas semipalmata), or the species was considered 'vagrant'. Using information from BirdLife Australia [41], the Atlas of Living Australia [42] and Australian Bird Data Version 1.0 [43], the feeding preferences of species were categorised from most preferred to occasional.
To give a general overview of any assemblage changes, the records of 1998 and 2018 were categorised into the following primary food sources: fruit, insects, invertebrates, nectar, omnivore, raptor, scavenger, seed, vegetation or vertebrate. If a species was considered to feed on two types of food source equally, both were considered the primary food source. Species were sorted by their level of specialisation: whether they had one, two, three or more food sources (Table A3).

Integration of Data
Following Hahs and McDonnell [44] and Conole and Kirkpatrick [30], the final edited spatial images were re-imported into ArcGIS and a 1 × 1 km grid was overlaid. The modified Bird Atlas data from 1998 and 2018 were then added as point layer files. As with Conole and Kirkpatrick [30], the locations were taken from the coordinates provided. The grid cells that contained records from both 1998 and 2018 were extracted for each year, and the level of land use type in each grid cell was calculated. Land use types were combined and simplified into greenspace (woodland, grass and forest), coastal (mangrove, sea and sand), urbanised (suburban, periurban, building and road), water and bare earth. Percentage change of greenspace and urbanisation between 1998 and 2018 was calculated for the relevant grid cells.

Statistical Analysis
Data were analysed using R version 3.6.1 [45]. Analysis of variance (ANOVA) was used to investigate whether the site and year had an effect on overall species numbers; paired t-tests were then used to examine the significance of any changes between 1998 and 2018 in those sites where urbanisation had increased or greenspace had decreased. The analysis was repeated to assess species' primary food source categories.
Finally, effects of land change on feeding specialisation numbers in bird assemblages were also explored. To do this, assemblage species proportions of one, two, three or more food sources were analysed for grid cells that were determined to have had an increase in urbanisation, a decrease in urbanisation, no change in the amount of urbanisation, an increase in greenspace or a decrease in greenspace.

Spatial Data
Results of the GEOBIA found that urbanised areas increased from 114 to 151 km 2 , a percentage increase of almost 3% of the total land mass measured. Greenspace, too, increased by just under 5%; 986 to 1045 km 2 .
For the 1998 imagery, overall accuracy was 75% (producer accuracy 72%; user accuracy 84%). The 2018 imagery returned an overall accuracy of 87% (84% producer and 90% user accuracies). This difference in accuracy was expected due to the quality differences in the Landsat 5 versus Landsat 8 imagery. Final corrections were made to the classified imagery before extraction of the relevant land use data for analysis ( Figure 2). creased or greenspace had decreased. The analysis was repeated to assess species' primary food source categories.
Finally, effects of land change on feeding specialisation numbers in bird assemblages were also explored. To do this, assemblage species proportions of one, two, three or more food sources were analysed for grid cells that were determined to have had an increase in urbanisation, a decrease in urbanisation, no change in the amount of urbanisation, an increase in greenspace or a decrease in greenspace.

Spatial Data
Results of the GEOBIA found that urbanised areas increased from 114 to 151 km 2 , a percentage increase of almost 3% of the total land mass measured. Greenspace, too, increased by just under 5%; 986 to 1045 km 2 .
For the 1998 imagery, overall accuracy was 75% (producer accuracy 72%; user accuracy 84%). The 2018 imagery returned an overall accuracy of 87% (84% producer and 90% user accuracies). This difference in accuracy was expected due to the quality differences in the Landsat 5 versus Landsat 8 imagery. Final corrections were made to the classified imagery before extraction of the relevant land use data for analysis ( Figure 2).

Bird Atlas Data
Overall, there were 23 grid cells that contained Bird Atlas records from both 1998 and 2018. Of these, nine showed an increase in urbanisation ranging from 0.9 to 58.2%; eight indicated a decrease in urbanisation and six were recorded as no change. A decrease in greenspace was found in 15 grid cells, and 8 were determined to have had an increase. No species considered primarily a scavenger was found in the 23 grid cells.

Bird Atlas Data
Overall, there were 23 grid cells that contained Bird Atlas records from both 1998 and 2018. Of these, nine showed an increase in urbanisation ranging from 0.9 to 58.2%; eight indicated a decrease in urbanisation and six were recorded as no change. A decrease in greenspace was found in 15 grid cells, and 8 were determined to have had an increase. No species considered primarily a scavenger was found in the 23 grid cells.

Overall Change in Species Numbers
In total, 165 individual species were represented in the bird records used. Of these, 23 species were recorded in 1998 only and 19 in 2018 only (Table A4). All sites showed a change in species numbers between 1998 and 2018; however, when all sites were included, there was no significant change in total species numbers (F = 1.1909, p = 0.3999 and F = 1.9395, p = 0.1707, respectively).
Focussing on those sites where urbanisation had increased or greenspace had decreased over the 20-year period, changes in species numbers were more distinctive. Al-though sites are listed numerically, this is for brevity, and they are not necessarily the same location.
Where urbanisation had increased, of the nine sites analysed, five sites showed an increase in overall species numbers with four of these increases found to be statistically significant (p < 0.05). Examination of sites where species numbers decreased found two where the decrease was significant and two where there was no statistical significance (Table 1). In sites where there was a decrease in greenspace, the majority (10 out of 15) displayed an increase in species numbers, but of these, only three were found to be significant. The remaining five sites where a decrease in species was recorded again found three with statistical significance (Table 2).

Changes to Species Numbers with Regard to Primary Food Sources
Species for each site were categorised by their primary food source, and changes in numbers initially were record as either an increase, decrease or no change for each food source.
In areas where urbanisation had increased, site 6 (increase in urbanisation of 5.8%) showed an increase in every species type, whereas site 4, which had the highest increase in urbanisation of any site (58.2%), displayed a decrease in almost every species type bar one, in which there was no change (Table 3). Fruit I  I  I  D  D  I  I  D  I  Insect  I  I  I  D  I  I  I  D  I  Invertebrate  I  D  I  D  D  I  D  D  D  Nectar  I  I  I  D  NC  I  I  NC  I  Omnivore  I  I  NC  D  NC  I  D  D  D  Raptor  I  I  I  D  D  I  D  NC  I  Seed  I  I  I  D  D  I  I  D  I  Vegetation  D  I  NC  NC  NC  I  NC  NC  D  Vertebrate  I  I  I  D  D  I  D NC NC Subsequent analysis found that there was no significant change in species numbers by primary food source overall (Table 4). Fruit I  I  D  I  I  D  I  D  I  D  NC  I  I  D  I  Insect  I  I  I  I  I  D  I  D  I  D  D  I  I  I  I  Invertebrate  I  I  I  I  I  D  I  D  D  D  I  D  D  I  I  Nectar  I  NC  I  D  I  D  I  D  I  NC  D  I  I  I  I  Omnivore  I  I  NC  D  I  D  I  D  D  D  I  D  D  NC  I  Raptor  I  I  I  D  I  D  I  D  D  NC  D  I  I  I  I  Seed  I  I  I  D  I  D  I  D  I  D  NC  I  I  I  I  Vegetation  D  NC  D  NC  I  NC  I  NC  NC  NC  D  D  D  NC  NC  Vertebrate  I  D  D  D  I  D  I  D  D  NC  NC  I  NC  NC  I When investigating those sites where greenspace had decreased, one site (7) showed an increase in every species type despite having a greater reduction in greenspace (14.2%) than either site 6 (13.4%) or site 8 (5.3%), both of which showed a reduction in every species type except those that feed primarily on vegetation (Table 4).
Unlike those sites where urbanisation had increased, analysis of changes in species numbers in the sites where greenspace had decreased found a significant change in those species whose primary food source was insects (p < 0.05). Some change was noted for raptors too, but this was not quite statistically significant (p = 0.06) ( Table 5).

Changes to Bird Assemblages with Regard to Primary Food Sources
Alongside changes to the number of species within bird communities over time, it was also pertinent to investigate whether the proportion of feeding guild types was altered within assemblages due to changing habitats, specifically in those sites where urbanisation increased or greenspace decreased.
A comparison of the proportions of species type for each site where urbanisation was found to have increased is shown in Figure 3.

Changes to Bird Assemblages with Regard to Primary Food Sources
Alongside changes to the number of species within bird communities over time, it was also pertinent to investigate whether the proportion of feeding guild types was altered within assemblages due to changing habitats, specifically in those sites where urbanisation increased or greenspace decreased.
A comparison of the proportions of species type for each site where urbanisation was found to have increased is shown in Figure 3. Subsequent analysis showed an increase in urbanisation resulted in no significant difference in the proportion of species of different primary food sources between 1998 and 2018.
However, in grid cells that had a decreased amount of greenspace, a slight significant difference was shown in those species that chiefly feed on insects, and there is a suggestion of change in those species that primarily feed on fruit, although this was not found to be significant (Table 6). A comparison of species proportion by primary food source for those grid cell sites where greenspace had decreased over the 20-year period is shown in Figure 4. A comparison of species proportion by primary food source for those grid cell sites where greenspace had decreased over the 20-year period is shown in Figure 4.

Levels of Specialisation
The levels of feeding specialisation within bird communities between 1998 and 2018 showed no significant change despite varying levels of urbanisation and greenspace (Table 7).

Levels of Specialisation
The levels of feeding specialisation within bird communities between 1998 and 2018 showed no significant change despite varying levels of urbanisation and greenspace (Table 7).

Discussion
Of the 258 birds species recorded in the Darwin area (escaped or introduced species and vagrants excluded), approximately 40% are considered resident [46] with the remaining species described as mobile. It has previously been thought that mobile species cope better with urbanisation than sedentary species [47,48], although other studies have suggested that highly mobile species may be more vulnerable to habitat changes due to their dependence on larger habitat patches and generally far-reaching home ranges [49][50][51]. Furthermore, there is also evidence to suggest that the fragmentation of habitats caused by urbanisation is going some way to changing the status of some birds from 'visitor' to 'resident' [52][53][54].
That there is a significant difference, albeit slight, in the proportion of those species that primarily feed on insects in areas where greenspace has decreased may possibly be attributed to the establishment of urban gardens. Although the size of the greenspace has decreased overall in these locations, in 73% of the grid cells analysed, the proportion of species that preferred insects increased; in contrast, only 47% of grid cells contained an increase in fructivores. It may be that the plants in these gardens attract a larger array of insects, thus providing more appealing spaces for insectivores. Numerous studies have indicated that the bird species that tend to thrive in urban environments are those that nest in cavities or canopies (sites less likely to be disturbed by human activity) and are more likely to be granivorous or omnivorous, whereas those species that avoid urban areas are those that predominantly nest in shrubs or trees or at ground level and have a more specialised diet, frequently insectivores [29,30,55]. However, these characteristics have been determined from research undertaken in either the Northern Hemisphere or the temperate zones of the Southern Hemisphere, places where garden vegetation is often vastly different to that of the original habitat. Suburban gardens of the Darwin region often contain native plant species alongside exotics, and this mixture may be enough to maintain, or even increase, insect numbers.
That there is no significant difference in the levels of feeding specialisation in bird assemblages is something of a surprise, as it would be fair to assume that more generalised species would be increasing and those species with one preferred food source would be forced out, particularly in areas of urban increase. Homogenisation of habitat is an oft-quoted result of urbanisation [5,56,57], but from this broad overview, it appears not to be affecting the composition of bird communities in the Darwin region. Further research into the characteristics of species making up urban assemblages will go some way to better understanding why this may be.
Invasion by feral species is a frequent result of an increased human presence in an area [29,30,58]. Many bird species that are unable to adapt to the suddenly changed environment will often move out of an area resulting in diminished biodiversity and allowing invasive species to move in [4,8,9]. Other species can benefit from urbanisation; these are often birds with a more generalist diet and life-history traits that are conducive to living in a fragmented habitat [10].
The only feral species in the records was Columba livia-the common pigeon, rock dove or rock pigeon. This species was recorded in the 1998 records only, most likely due to an eradication programme undertaken by the Parks and Wildlife Commission, NT, beginning in 1996, that saw the species considered eradicated from the Darwin area by 2004 [59,60]. The lack of established feral bird populations in the Darwin region could, potentially, suggest that native bird species are filling the 'urban exploiter' or 'suburban adapter' niche that is commonly filled by exotic species in other cities, but there is little or no research to show what these species may be. Additionally, a unique aspect of the Darwin region is that it is the only capital in Australia where urban fires are a regular annual occurrence. The city is located within the monsoonal tropics; a landscape dominated by tropical savannas, experiencing frequent fires during the prolonged dry season from May to October, often within just metres of suburban housing areas [61]. In addition to wildfires, many of these fires are prescribed, being undertaken by various land-management bodies as part of a yearly landscape management approach [62,63]. Whether these annual fires are in some way supporting and maintaining the local bird assemblages whilst preventing invasive species from establishing populations is an intriguing question. Research into the combined impact of urban fires and human modification of the environment on local bird populations over time is ongoing; it is hoped that the results will go some way towards answering this question.

Conclusions
Whilst very transient, the population of Darwin and its surrounds has grown significantly between the years 1998 and 2018, and urbanisation has expanded more than 50 km from the city centre. Despite this, this general overview of bird assemblages shows that although species numbers fluctuate, avian communities appear to be stable. The increase in the insectivorous species may be due to the establishment, maturity and vegetative composition of suburban gardens. This may potentially signify a shift in the species diversity of the area; however, further research is needed to investigate this.

Acknowledgments:
The authors would like to thank Andrew Silcox and BirdLife Australia for access to bird data records.

Conflicts of Interest:
The authors declare no conflict of interest.