Woodland birds are a species assemblage of conservation concern, and their persistence in fragmented agricultural landscapes is dependent on both the preservation of existing woodland remnants and the implementation of restoration plantings. However, little is known about the habitat-use and persistence of birds in fragmented agricultural landscapes. We present a detailed, population-oriented study of woodland birds in temperate eucalypt woodland restoration plantings and remnant woodland patches in the South-west Slopes bioregion of New South Wales, Australia. First, we undertook a three-year mark-recapture project to assess annual survival and site fidelity in restoration plantings and woodland remnants. We supplemented our recapture efforts with resightings of colour-banded individuals. Second, we tracked individual birds of two species – superb fairywren (Malurus cyaneus) and willie wagtail (Rhipidura leucophrys) – and documented snapshots of their home ranges and movement patterns during the breeding season. Annual survival in the woodland bird assemblage was lower than expected (51%). Home ranges of the superb fairywren were positively correlated with patch size, and were constrained by patch edges in linear sites. Superb fairywrens and willie wagtails were more likely to travel longer distances between substrates while foraging in linear sites. Willie wagtails engaged in significant gap-crossing (up to 400 m) between adjacent habitat patches. Our findings indicate that 1) patch isolation and certain patch configurations place resident birds at an energetic disadvantage, and 2) in our study area, woodland bird populations are continuing to decline. We recommend landscape-scale restoration programs aim to address ongoing population declines. Studies such as ours conducted over longer time periods would provide a deeper understanding of habitat-use and population processes of woodland birds in fragmented agricultural landscapes.

Study sites

We used long-term bird survey data to select a subset of sites in which to conduct our study. We selected 12 restoration plantings and six similarly sized woodland remnants of varying size (1.3-7.7 ha) and shape (linear vs. block-shaped; calculated numerically as perimeter/width). We additionally chose three large (>47 ha) remnant patches of woodland to serve as reference sites, representing good quality woodland in the study area. We chose sites with low abundances of the noisy miner (Manorina melanocephala), a hyperaggressive native honeyeater that is known to exclude small woodland birds. Our sites were separated geographically by a minimum of 500 m to promote spatial independence. Restoration plantings were aged between 12 and 25 years, and were characterised by a Eucalyptus overstorey and an understorey of predominantly Acacia shrubs. Trees and shrubs were planted for ecological purposes, and were usually fenced for protection from grazing by livestock. Remnant woodland patches typically constituted a Eucalyptus overstorey plus an Acacia-dominated understorey of varying density. Remnant sites tended to contain more coarse woody debris (fallen branches and trees) than restoration plantings. All sites featured a ground layer that was usually dominated by exotic pasture grasses, with various amounts of leaf litter, native forbs, native grasses, weeds, bare ground, moss/lichens, rocks, and coarse woody debris.

Bird banding and recaptures

We undertook initial banding of woodland birds in July-October of 2015. Subsequent recapture and banding efforts took place in June-August of 2016 and 2017. We conducted an approximately equal number of net hours in each study site, with net sites selected based on suitable habitat features and observed passage routes to maximise capture rates. Once selected, we used the same net sites consistently throughout the three years of the study, except when vegetation growth or tree collapse prevented the use of existing net sites. In these cases, we moved nets to suitable locations as close as possible to the original net sites. We used two nets that were 6 m in length, and four nets that were 9 m in length. We banded over two consecutive days in each study site. On the first day, we operated nets for approximately four hours pre-sunset. On the subsequent day, we opened nets from half an hour pre-dawn to approximately four hours post-sunrise. Inclement weather sometimes required us to close nets, and occasionally prevented us from banding on consecutive days in our study sites. In these instances, we resumed banding at the earliest opportunity once the weather had cleared.

We banded birds with a standard metal band (aluminium or aluminium alloy) provided by the Australian Bird and Bat Banding Scheme. Ten species also received a combination of coloured plastic and/or metal leg bands. These were the superb fairywren (Malurus cyaneus), yellow-rumped thornbill (Acanthiza chrysorrhoa), buff-rumped thornbill (A. reguloides), grey shrikethrush (Colluricincla harmonica), rufous whistler (Pachycephala rufiventris), red-capped robin (Petroica goodenovii), willie wagtail (Rhipidura leucophrys), white-browed babbler (Pomatostomus superciliosus), speckled warbler (Pyrrholaemus sagittatus), and diamond firetail (Stagonopleura guttata). We released birds within 300 m of their initial capture site. Any juveniles captured were released as close as possible to their initial capture site. For the purposes of our study, the “woodland assemblage” included all species that were captured in our study sites.

Resightings

In the breeding season of 2015 (September-December), we recorded incidental sightings of colour-banded birds. In 2016, we undertook surveys to record sightings of colour-banded individuals. We undertook random area searches in each study site, with the length of time designated per unit area. We surveyed small sites (1.4 ha search area) for one hour, and large sites (3 ha search area) for two hours. A skilled observer searched a patch for woodland birds, and recorded colour-band combinations and GPS locations of colour-banded individuals when sighted. Sites were surveyed once per month from September to November. We also recorded incidental sightings of colour-banded individuals while visiting study sites throughout the breeding season.

The seasonally migratory rufous whistler had not yet returned to the study region when recapture efforts took place in 2016 and 2017. We resighted colour-banded individuals incidentally in 2015, and during thorough site surveys in 2016. In 2017, we undertook targeted surveys to identify whether colour-banded rufous whistlers had returned to their territories. We visited known territories of colour-banded males in October 2017 and used call-playback for up to 5 minutes to elicit a territorial response. If a whistler responded, we immediately ceased playback and followed the bird until we could ascertain its band status/combination.

Home range tracking

We chose two target species for home range tracking – the superb fairywren, and the willie wagtail. Both species are relatively common in habitat patches in agricultural landscapes, and can use the matrix to some extent (usually foraging at the edges of patches). However, the two species differ in their movement patterns and habitat preferences. Additional reasons for choosing these two species included their bold nature, and their ease of detection (facilitating tracking of individuals).

Superb fairywrens were fitted with coloured leg bands for ease of tracking and identification. For consistency, we tracked only male fairywrens. Not all tracked willie wagtails had been banded. However, the low density of willie wagtails in the study sites, their preference for open habitats, and the presence of individually distinct plumage or feather moult characteristics enabled us to reliably track individual birds over the course of several hours.

We selected birds to track at random. Upon arriving at a study site, an observer would locate an individual (typically the first bird encountered) on which to focus. Based on a pilot study in 2015, we determined that a minimum of 30 home range points was required to map a representative snapshot of the home range of a male superb fairywren or a willie wagtail. We therefore followed birds until at least 30 points had been recorded. We followed birds for a maximum of four hours. We used a handheld Garmin eTrex GPS device, accurate to the nearest 3 m, to record home range data. Observers typically watched birds from a distance of around 20 m, using a pair of handheld binoculars, so as to not disrupt or influence the behaviour or movements of the birds. Observers also waited until a bird had moved away from each distinct substrate before approaching to record the GPS location.

Statistical analyses

Survival and site fidelity

To analyse recapture and resighting data, we used Program MARK via the R package “RMark” in R version 3.5.2. We used the Cormack-Jolly-Seber (CJS) model to estimate apparent survival for the entire assemblage. For colour-banded birds, and for the superb fairywren, we also used the Barker model to estimate survival and site fidelity. This model allows the inclusion of ad hoc resightings of colour-marked individuals, separates permanent emigration from true mortality, and relaxes the assumption of no permanent emigration.

Home ranges and movement analyses

We used ArcMap Desktop version 10.6.1 to plot GPS coordinates of home ranges. We used the Minimum Bounding Geometry function to calculate 100% Minimum Convex Polygons (MCPs) for all home ranges. We also calculated the perimeter and width of each home range. Additionally, we calculated the distance moved by individual birds between distinct substrates (GPS points) during a tracking period, and documented “gap-crossing” behaviour. We defined “substrates” as distinct units of vegetation (individual trees, shrubs), woody debris (logs, fallen branches), artificial substrates (fence wires, posts), and patches of ground on which a bird perched or foraged. We defined gaps as spaces between patches of native vegetation. A “patch” of native vegetation may comprise a single paddock tree.

0 = not recorded

1 = recorded