Get 20M+ Full-Text Papers For Less Than $1.50/day. Start a 14-Day Trial for You or Your Team.

Learn More →

A bird occupancy estimator for land practitioners in the NSW South Western Slopes bioregion

A bird occupancy estimator for land practitioners in the NSW South Western Slopes bioregion IntroductionBiodiversity loss is a major global issue (Ceballos et al. 2017; Intergovernmental Science‐policy Platform on Biodiversity and Ecosystem Services (IPBES) 2019), and a significant amount of the past biodiversity loss has occurred in agricultural landscapes, both internationally (Maxwell et al. 2016; Filazzola et al. 2020) and in Australia (Williams and Price 2011). Globally, there have been major restoration efforts to recover biodiversity in agricultural landscapes (Brancalion & Holl 2020; Chazdon et al. 2020; Crouzeilles et al. 2020). However, greater information on the biodiversity dividends from restoration actions is often desired, including by on‐the‐ground practitioners responsible for implementing restoration programmes (Munro and Lindenmayer 2011). Such information is particularly important in extensively modified ecosystems such as the endangered Box Gum Grassy Woodland ecosystems, which characterise much of the wheat–sheep belt in northern Victoria, inland New South Wales and south‐east Queensland (Hobbs and Yates 2000). In these ecosystems, there have been long‐held concerns about declines in temperate woodland birds (Lindenmayer et al. 2018c; Ford et al. 2001). There also has been a need for information on the effectiveness of interventions such as plantings to reverse these declines (Belder et al. 2018; Lindenmayer et al. 2018b).We have developed a new, free webtool, BirdCast [https://sustfarm.shinyapps.io/BirdCast/], that has potential to describe the possible bird biodiversity dividends generated by investments in vegetation management and restoration programmes. BirdCast estimates the response of bird biodiversity in Box Gum Grassy Woodland to both existing woody vegetation and possible future woody vegetation after revegetation or natural regeneration in the NSW South Western Slopes bioregion (which also extends into northern Victoria) (DAWE 2012). We anticipate that BirdCast will be used by interested farmers to inform them about their farm's potential biodiversity and provide greater context to environmental management decisions. We anticipate that other natural resource management professionals will use BirdCast when engaging with farmers on management topics or decision‐making. BirdCast can produce estimates of:The chance of occupancy (occupancy probability) in one or more user‐specified woodland areas by 60 different bird species, including five species of conservation concern: Brown Treecreeper, Diamond Firetail, Dusky Woodswallow, Grey‐crowned Babbler and Superb Parrot (Department of Environment, Energy and Science 2021).Avian species richness.Bird biodiversity dividends from some management interventions, including:Addition or removal of nearby woody vegetation cover (i.e. vegetation restoration, natural regeneration or land clearing).Addition of new areas of planted eucalypt woodland.Interventions that prevent Noisy Miner occupancy of a woodland area.The webtool also builds user‐friendly key resources. These include the following:Reports for printing.Downloadable tables of estimates.Visualisations of estimates.A compilation of engaging bird images and descriptions (from BirdLife Australia and BirdLife Australia photographers) for the birds most likely to inhabit Box Gum Grassy Woodlands in farming landscapes.BirdCast is underpinned by robust statistical relationships developed from analyses of 17 years of bird observations (e.g. see Lindenmayer et al. 2018c, 2020) from 518 sites in temperate woodlands, primarily in New South Wales. BirdCast's quantitative estimates for scenarios at the scale of individual or multiple woodland areas make it a unique resource. Other resources typically provide species descriptions, species distribution maps or information on previous observations (Campbell et al. 2015; Atlas of Living Australia 2022; Shelley et al. 2022). These resources either do not have the quantitative depth of BirdCast or do not easily allow comparisons of biodiversity response to different management scenarios.This paper is an introduction to BirdCast for restoration practitioners and natural resource managers. We provide some background to the models and data used to create BirdCast (A Joint‐Species Statistical Model). The webtool was developed from the statistical model through user consultations and expert digital designers (Development of the Webtool). We then provide an example (A Worked Example), guidance on the interpretation of BirdCast outputs (Interpreting the Outputs of BirdCast) and limitations (Limitations).A Joint‐Species Statistical ModelThe first stage in the development of BirdCast (Fig. 1, left) was a statistical model for the occupancy of bird species in individual woodland areas. We used 5189 expert bird surveys in 518 different areas of remnant Box Gum Grassy Woodland or (re)planted woodland to create this model. These woodland areas were located primarily in the NSW South Western Slopes bioregion (DAWE 2012), with a lesser number of areas at locations up to the Queensland–NSW border (Fig. 2). The woodland areas were typically between 2 ha and 10 ha in area. For each woodland area, a representative survey site, 200 m long and 100 m wide, was selected. There was no minimum distance between different woodland areas, and more than 60% of sites were within 1 km of another site. The surveys were conducted in spring, spanning a time frame of 17 years. Surveys were generally repeated at least twice in a season to account for detection in the statistical model (Cunningham et al. 1999; Field et al. 2002). Planted woodlands were dominated by eucalypts 3+ years old, fenced at the time of planting and established using tubestock or direct seeding.1FigureDevelopment of BirdCast. Left: development of the statistical model. Right: development of the webtool.2FigureLocations of survey sites. Plus symbol (+): sites in remnant Box Gum Grassy Woodland. Green dots: sites in compatible planted woodland. Shaded region: NSW South Western Slopes bioregion (DAWE 2012). [Colour figure can be viewed at wileyonlinelibrary.com]The model included 62 species that were detected in 100 or more bird surveys. These species were mostly sedentary and land‐based. Five of the species were of conservation concern: Brown Treecreeper, Diamond Firetail, Dusky Woodswallow, Grey‐crowned Babbler and Superb Parrot (Department of Environment, Energy and Science 2021). The model accounted for interspecies correlation and imperfect detection (Tobler et al. 2019). Our method for selecting environmental predictors of species occupancy followed Hingee et al. (2022) and is the topic of the Supporting Information (Appendix S1–S3). The model passed a number of diagnostic tests (Section 5.5 of Appendix S1: Model fitting and selection), suggesting that estimates of occupancy probability, and estimated uncertainty of these estimates, can be trusted when the modelling assumptions are satisfied. We saw no indication of spatial correlation in the model's residuals for occupancy, despite the model having no spatially explicit dependence component. From our diagnostics, we suspect that much of the spatial correlation was taken on by the model's latent variables.Major site‐scale predictors of species occupancy were the presence of the hyper‐aggressive native honeyeater, the Noisy Miner (Manorina melanocephala), and whether the woodland area was planted versus remnant. Important covariates related to the surrounding landscape were the amount of woody vegetation cover within 500 m of the site centre and further away (within 3 km of the site centre), both of which were derived from satellite information (Liao et al. 2020). These woody vegetation cover data, sometimes termed foliage cover, were estimates of the proportion of land covered by 2 m + high branches or foliage (Liao et al. 2020). Australia's temperate woodlands typically have between 10% and 70% woody vegetation cover, depending on the age and structure of the vegetation (Hnatiuk et al. 2010). The woody vegetation cover estimates derived from satellite data may also fluctuate with water availability (Liao et al. 2020). Maps of the amount of woody vegetation cover can be viewed at [http://anuwald.science/tree], and a tool for obtaining the woody cover percentages within 500 m and 3 km of any location was integrated into BirdCast. Climatological averages were also crucial predictors of bird occupancy.Development of the WebtoolUsing R shiny (R Core Team, 2020; Chang et al. 2021), we built an initial webtool. In this version, important predictors and model outputs were shown on the same page. The user was required to input woody cover percentages limited to between 2% and 20%, which was approximately 90% of the range of woody cover percentages used to build the statistical model. We removed two species that risked confusing users: the Australian Wood Duck relies on waterbodies, so we expect the model estimates to be misinformative; and interpretation of occupancy probability estimates for the Australasian Pipit is prone to confusion as Australasian Pipits are much more likely to inhabit open farm paddocks than woodlands. The remaining 60 species in the webtool are listed in the Appendix A. Users were able to compare estimates with a previously saved scenario through graphics showing the relative increase or decrease (i.e. ratio) in species occupancy probability.When the user specified multiple woodland areas, the occupancy probability was defined as the chance that the species occupied at least one of the specified woodland areas. This form of occupancy probability was estimated by assuming that a species occupying one woodland area would also occupy any woodland areas that were more favourable according to our statistical model. This could be considered as a lower bound on the true occupancy probability – weaker statistical dependence between woodland areas will lead to higher occupancy probability estimates. Mathematically, the estimate for each species was the maximum of the occupancy probability estimate for each individual woodland area. This maximum is simple to compute from the results of our statistical model and produces higher values than an average of the estimates.This initial version was demonstrated to landholders, Local Land Services staff and other Natural Resource Management professionals at workshops in Orange and Wagga Wagga. A third workshop, scheduled for Wangaratta, was postponed due to COVID‐19. Detailed feedback on the webtool was also sought through consultations with potential users.We then engaged expert digital designers for further feedback and to design a new version of the webtool based on all prior feedback. This design had an easy linear workflow for users, was more touchscreen‐friendly, made comparisons easier to use, had cleaner graphs and moved engaging BirdLife Australia Photography images into the user's main view. We also incorporated interactive maps (Cheng et al. 2021) so that users could load satellite‐based estimates of woody cover by clicking on the centre of their woodland area.BirdCast can be used by visiting [https://sustfarm.shinyapps.io/BirdCast/] with any modern Internet browser on a desktop, tablet or phone. A demonstration video of BirdCast is available in BirdCast's user guide, accessed at the top of the webtool. Users will find it helpful to have some experience in using modern online interfaces, including online maps and satellite imagery. A live Internet connection is required. The source code for BirdCast is available at [https://github.com/sustainablefarms/farm_biodiversity_app/] and can be used with RStudio (RStudio Team 2020) to run BirdCast offline without the BirdLife Australia Photographs. BirdCast uses R shiny (R Core Team, 2020; Chang et al. 2021) and several packages written in the statistical programming language R (Wickham 2016; Pebesma 2018; Sievert 2020; Xie et al. 2020; Cheng et al. 2021; Coene 2021).A Worked ExampleStep 1: Your farmSuppose we are interested in a remnant area of Box Gum Grassy Woodland in the Cowra region of NSW. Noisy Miners occupy this area, and it has 9.5% and 7.5% woody vegetation cover within 500 m and 3 km, respectively, of the area's centre (these values were the average values for our survey sites in the Cowra region during the years that bird surveys were conducted). We provide all this information in Step 1 of BirdCast to create Scenario 1 (Fig. 3). If we knew the GPS co‐ordinates for the location of the woodland area, then BirdCast could obtain the woody vegetation cover estimates from satellite data (Liao et al. 2020).3FigureIn Step 1 of BirdCast, the user provides their region (left) and attributes of their woodland areas (right). This information becomes Scenario 1 in BirdCast. [Colour figure can be viewed at wileyonlinelibrary.com]Step 2: Bird diversityIn Step 2, BirdCast estimates the occupancy probability of bird species for the woodland area (Fig. 4). We can see that, on average, approximately 12 of the 60 species in BirdCast are estimated to occupy the woodland area in Scenario 1. In comparison, the average woodland area in our data set is estimated to have approximately 14 species. If woody vegetation cover within 500 m of the centre of the Scenario 1 woodland area was equal to 2%, then BirdCast estimates that there would be approximately 1 fewer species occupying the woodland on average. If the woody vegetation cover within 500 m was equal to 20%, then BirdCast estimates that there would be approximately 2 additional species occupying the woodland area on average.4FigureIn Step 2, BirdCast presents occupancy probability estimates for Scenario 1. The top display shows estimates of expected species richness under Scenario 1, in an average woodland area in our data set, and under two variations to Scenario 1. Other displays show the most likely and least likely species according to the occupancy probability estimates, details for the five vulnerable species in BirdCast, and a figure of occupancy probability estimates for all species in BirdCast. [Colour figure can be viewed at wileyonlinelibrary.com]Step 2 also includes displays of the 10 most likely species and 10 least likely species to occupy the woodland area and the five vulnerable species in BirdCast. These displays include engaging images from BirdLife Australia Photography, with further information displayed when clicking on an image. A final display shows the occupancy probability estimate for every species in BirdCast, and a table of these estimates is available for download. A full pdf report of Step 2 can be downloaded at the bottom of the page.Steps 3 and 4: Create a new scenario and compare bird diversityUsing BirdCast, we can view the estimates of bird species likely to occupy woodland areas and compare them with the species we would expect to see given changes to some key habitat and landscape attributes – such as fewer nearby trees, a new planting and an absence of Noisy Miners (which represents an improvement in habitat condition).No Noisy MinersIn Step 3, we create a second scenario, Scenario 2. BirdCast helpfully prepopulates Scenario 2 from the information in Scenario 1. For Scenario 2, suppose Noisy Miners are absent. We select ‘no’ in answer to the question about Noisy Miner presence for the woodland area, and then save and close the woodland area window. Removing Noisy Miners from a woodland area, or removing their impact, may be possible by increasing midstorey (Lindenmayer et al. 2018a). We then move to Step 4, which compares bird diversity between Scenario 1 and Scenario 2. BirdCast estimates that removing Noisy Miners increases average species richness by approximately 1 species and that the 10 most likely species now include the Black‐faced Cuckoo‐shrike and Willie Wagtail (Fig. 5). From the vulnerable species display, we can see that BirdCast estimates that there is a 26% chance of the Brown Treecreeper occupying the woodland area in Scenario 2 (without Noisy Miners), and only a 10% chance in Scenario 1 (with Noisy Miners). From the relative occupancy probability display, we can see that BirdCast estimates that removing Noisy Miners improves the occupancy chances for nearly all smaller‐bodied species, especially the Brown Thornbill, Leaden Flycatcher and Silvereye.5FigureIn Step 4, BirdCast compares bird species occupancy estimates for Scenario 1 and Scenario 2. The displays are similar to the displays in Step 2. [Colour figure can be viewed at wileyonlinelibrary.com]Removal of woody vegetation coverWe go back to Step 3. Suppose Noisy Miners are present again and some trees on the edge of the woodland area are removed so that we reduce woody vegetation cover within 500 m of the woodland's centre to 7.5% (from an original value of 9.5%). We then move forward again to Step 4 of BirdCast to compare Scenario 1 with this new Scenario 2. BirdCast estimates that the change in woody vegetation cover between Scenario 1 and Scenario 2 slightly lowers the average species richness, although the average richness remains approximately 12 species. In the relative occupancy section, we can see that most species are less likely to occupy the reduced woodland area (Scenario 2), especially small‐bodied bird species, although a few species are more likely to occupy the woodland area, such as the Brown Songlark and the Tree Martin, which are commonly associated with more open or cleared landscapes.A new woodland areaIf a new 1 ha woodland area (e.g. a 100 m × 100 m block planting) is established close to the first woodland area, then this creates a new type of habitat (planted woodland) and increases the woody vegetation cover near the first woodland area. Based on typical woody vegetation cover fractions for remnant temperate woodlands, the new woodland area could eventually result in 0.4 ha of additional woody vegetation cover, which increases the woody vegetation cover near the first, remnant, woodland area to 10% (from an original of 9.5%). We assume Noisy Miners are present in both. In Step 3, for the existing (remnant) woodland area we set Noisy Miners as present and 10% nearby woody vegetation cover. We then add a woodland area, select ‘planted woodland’ and set it to have 10% woody vegetation cover with 500 m. In Step 4, BirdCast suggests that the additional woodland area increases the expected richness of the farm by approximately 3 species. Furthermore, BirdCast estimates that Yellow Thornbills, Rufous Whistlers, and Red Wattlebirds are each more than three times as likely to occupy at least one of the woodland areas in Scenario 2 compared with the single woodland area in Scenario 1.Interpreting the Outputs of BirdCastThe outputs of BirdCast are most reliable for woodland areas that are:Eitherremnant Box Gum Grassy Woodlands, or3+ years old eucalypt‐dominated plantings that were established by tubestock or direct seeding and fenced at the time of planting.In or near the South Western Slopes bioregion (DAWE 2012), roughly the lower inland slopes of the Great Dividing Range, from Benalla in Victoria to Dubbo in NSW.Within a modified agricultural landscape dominated by livestock grazing or mixed farming enterprises, that is a relatively cleared landscape.Additionally, the estimates of occupancy are for springtime and average climatic conditions, which characterised the climate of the South Western Slopes between 2000 and 2017.When multiple woodland areas are specified in BirdCast, then species occupancy probability estimates are for the occupancy of any of the areas. These woodland areas must be sufficiently close to each other that bird species can easily colonise more favourable patches. Furthermore, the estimates apply to the specified woodland areas, not the neighbouring paddocks.The relative occupancy probability for a given species is the ratio of two different estimates. The numerator is the estimate of occupancy probability for Scenario 2, and the denominator is the occupancy probability estimate for Scenario 1. The larger and more positive the ratio, the larger the improvement in occupancy probability over the reference scenario. For example, a relative occupancy probability of ‘3’ means that the species is estimated to be three times more likely to occupy the woodland area(s) in Scenario 2 than the woodland area(s) in Scenario 1.Accuracy of estimates for species occupancy probability, in the form of 95% highest posterior density intervals, can be shown as error bars in many of the displays in BirdCast. When the modelling assumptions are met (Section S4.1), there is a 95% chance that the true occupancy probability is within the error bars. This uncertainty is due to the uncertainty of the fitted values for the parameters in the statistical model. Inaccuracies induced by the maximum assumption for multiple woodlands (A Joint‐Species Statistical Model) are not included.The estimates of species richness in BirdCast should be viewed as estimates of the average number of common, land‐based bird species occupying the woodland area(s). This expected species richness was the sum of the occupancy probability estimates of all species in BirdCast. Due to the mathematical properties of statistical expectations, no threshold on occupancy probabilities was required. These expected species richness estimates are unlikely to be equal to actual avian species richness because, according to the statistical model for a single woodland area, the realised number of species will vary with a standard deviation of four from the estimated average. Corresponding error bars have not been included in BirdCast to reduce confusion and due to the computational burden of the species interaction components of the statistical model. Furthermore, the estimate ignores water birds and many rare bird species that were not included BirdCast (see the Appendix A for the list of bird species in BirdCast).LimitationsDue to the finite nature of our data, and the difficulty in jointly modelling bird species, there are many important situations that BirdCast cannot directly assist with. Due to the locations of our bird survey sites, BirdCast should not be used to estimate bird diversity outside the NSW South Western Slopes bioregion, within urban areas or in large nature reserves (e.g. national parks). Similarly, there are many rare and endangered species that are absent from BirdCast due to insufficient detection of these species. BirdCast does not directly estimate bird diversity given fencing, underplanting and different management of farm paddocks as there were insufficient data on these activities to include in our statistical model. Additionally, BirdCast estimates should not be used to estimate bird diversity after extreme events such as wildfires or long‐term climate change.When using BirdCast, it is also important to remember that the results do not account for:Interactions between species (apart from the impact of Noisy Miners on other bird species). Although the statistical model was able to incorporate some forms of interactions between species (Tobler et al. 2019), the computational burden was too high to include in BirdCast.Connectivity of landscapes, additional to the impact of nearby and regional woody vegetation cover – landscape connectivity may have little association with occupancy probability of bird species in this region (Lindenmayer et al. 2020).Presence of surface water (riparian waterways and dams).Randomness in the relationship between occupancy of nearby woodland areas. Our statistical model appears to have accounted for associations between nearby woodland areas through fitted latent variables (Section S5.5). However, these fitted latent variables do not easily allow for statistical prediction of these associations in unsurveyed woodland areas.Planned near‐term improvements to BirdCast include improved relative occupancy probability visualisations and faster loading of the interactive maps. There is potential for enabling estimates in regions where we have bird survey data (Fig. 1) further north of the NSW South Western Slopes bioregion. We are also particularly interested in improving BirdCast's estimates when multiple woodland areas are specified.ConclusionsBirdCast is a free webtool that provides farm‐specific bird occupancy estimates under different scenarios. It enables quantitative estimates of the bird species assemblages in response to some common restoration interventions and provides an opportunity to explore the effects of proposed interventions in an interactive model that can engage and motivate farmers. The tool may also be used as a prioritisation tool by natural resource management agencies to predict and direct restoration efforts to achieve the best bang for buck investment scenarios.AcknowledgementsWe acknowledge Dr Martin Westgate's crucial contributions to the development of BirdCast. Many valuable improvements were suggested by Suzannah Macbeth. Additional improvements were suggested by Colleen O'Malley, Michelle Young and early users of BirdCast. The authors thank the participants of our workshops and consultations, and the team at CRE8IVE. Quantities derived from spatial data played a crucial role in developing the model and were computed by Prof. Albert van Dijk. The authors acknowledge BirdLife Australia for the generous use of the species descriptions in Bird Cast and the following from BirdLife Photography for their photographs: Stephen Garth, Mark Lethlean, Michael Hamel‐Green, Glenn Pure, Jonathan Pridham, Edward Mikucki, Geoffrey Stapley, Lindsay Hansch, Woody Woodhouse, John Barkla, Kathy Zonnevylle, Richard Smart, Rob Parker, Trevor Bullock, Shane Baker, Ian Wilson, Chris Dubar, Anthony Thompson, Brian O'Leary, Bill Carroll and Con Boekel. Open access publishing facilitated by Australian National University, as part of the Wiley ‐ Australian National University agreement via the Council of Australian University Librarians.ReferencesAtlas of Living Australia (2022) Explore Your Area. CSIRO, Canberra, Australia. [Accessed 13 Feb 2022]. Available from URL:. https://biocache.ala.org.au/explore/your‐area.Belder D. J., Pierson J. C., Ikin K. and Lindenmayer D. B. (2018) Beyond pattern to process: current themes and future directions for the conservation of woodland birds through restoration plantings. Wildlife Research 45 (6), 473–489. https://doi.org/10.1071/WR17156Brancalion P. H. S. and Holl K. D. (2020) Guidance for successful tree planting initiatives. Journal of Applied Ecology 57 (12), 2349–2361. https://doi.org/10.1111/1365‐2664.13725.Campbell I., Woods S. and Leseberg N. (2015) Birds of Australia: A Photographic Guide. Princeton University Press, Princeton, New Jersey.Ceballos G., Ehrlich P. R. and Dirzo R. (2017) Biological annihilation via the ongoing sixth mass extinction signaled by vertebrate population losses and declines. Proceedings of the National Academy of Sciences of the United States of America 114 (30), E6089–E6096. https://doi.org/10.1073/pnas.1704949114.Chang, W., Cheng, J., Allaire, J. J. et al. (2021). Shiny: web application framework for R. https://CRAN.R‐project.org/package=shinyChazdon R. L., Lindenmayer D., Guariguata M. R., Crouzeilles R., Rey Benayas J. M. and Chavero E. L. (2020) Fostering natural forest regeneration on former agricultural land through economic and policy interventions. Environmental Research Letters 15, 43002. https://doi.org/10.1088/1748‐9326/ab79e6.Cheng, J., Karambelkar, B., & Xie, Y. (2021). Leaflet: create interactive web maps with the JavaScript “leaflet” library. https://CRAN.R‐project.org/package=leafletCoene, J. (2021). Waiter: loading screen for “shiny.” https://CRAN.R‐project.org/package=waiterCrouzeilles R., Beyer H. L., Monteiro L. M. et al. (2020) Achieving cost‐effective landscape‐scale forest restoration through targeted natural regeneration. Conservation Letters 13 (3), e12709. https://doi.org/10.1111/conl.12709.Cunningham R. B., Lindenmayer D. B., Nix H. A. and Lindenmayer B. D. (1999) Quantifying observer heterogeneity in bird counts. Australian Journal of Ecology 24, 270–277.(DAWE) Dept of Agriculture, Water and the Environment. (2012). Interim Biogeographic Regionalisation for Australia (IBRA), Version 7 (Subregions)—States and Territories. www.environment.gov.au/fed/catalog/search/resource/details.page?uuid={1273FBE2‐F266‐4F3F‐895D‐C1E45D77CAF5}Department of Environment, Energy and Science (2021) Species Found in the Grassy Woodlands Formation. New South Wales Department of Environment, Energy and Science, Australia. https://www.environment.nsw.gov.au/threatenedSpeciesApp/VegSpecies.aspx?vegName=Grassy+woodlands&habitat=F.Field S. A., Tyre A. J. and Possingham H. P. (2002) Estimating bird species richness: how should repeat surveys be organized in time? Austral Ecology 27, 624–629.Filazzola A., Brown C., Dettlaff M. A. et al. (2020) The effects of livestock grazing on biodiversity are multi‐trophic: a meta‐analysis. Ecology Letters 23, 1298–1309. https://doi.org/10.1111/ele.13527.Ford H. A., Barrett G. W., Saunders D. A. and Recher H. F. (2001) Why have birds in the woodlands of southern Australia declined? Biological Conservation 97 (1), 71–88. https://doi.org/10.1016/S0006‐3207(00)00101‐4.Hingee, K. L., Westgate, M. J. & Lindenmayer, D. B. (2022). Long‐Term Monitoring in Endangered Woodlands Shows Effects of Multi‐Scale Drivers on Bird Occupancy Journal of Biogeography, 49 (5). https://doi.org/10.1111/jbi.14353.Hnatiuk R. J., Thackway R. and Walker J. (2010) Vegetation. In: National Committee on Soil and Terrain, Australian Soil and Land Survey Field Handbook. CSIRO Publishing, Victoria, Australia. https://www.publish.csiro.au/book/5230.Hobbs R. J. and Yates C. J. (2000) Temperate Eucalypt Woodlands in Australia: Biology, Conservation, Management and Restoration. Surrey Beatty & Sons, NSW, Australia.Intergovernmental Science‐policy Platform on Biodiversity and Ecosystem Services (IPBES) (2019) IPBES Global Assessment Summary for Policymakers. United Nations.Liao Z., Van Dijk A. I. J. M., He B., Larraondo P. R. and Scarth P. F. (2020) Woody vegetation cover, height and biomass at 25‐m resolution across Australia derived from multiple site, airborne and satellite observations. International Journal of Applied Earth Observation and Geoinformation 93, 102209. https://doi.org/10.1016/j.jag.2020.102209.Lindenmayer D. B., Blanchard W., Crane M., Michael D. and Florance D. (2018a) Size or quality. What matters in vegetation restoration for bird biodiversity in endangered temperate woodlands? Austral Ecology 43 (7), 798–806. https://doi.org/10.1111/aec.12622.Lindenmayer D. B., Blanchard W., Crane M., Michael D. and Sato C. (2018b) Biodiversity benefits of vegetation restoration are undermined by livestock grazing. Restoration Ecology 26, 1157–1164.Lindenmayer D. B., Blanchard W., Foster C. N. et al. (2020) Habitat amount versus connectivity: an empirical study of bird responses. Biological Conservation 241, 108377. https://doi.org/10.1016/j.biocon.2019.108377.Lindenmayer D. B., Lane P., Westgate M. et al. (2018c) Tests of predictions associated with temporal changes in Australian bird populations. Biological Conservation 222, 212–221. https://doi.org/10.1016/j.biocon.2018.04.007.Maxwell S., Fuller R. A., Brooks T. and Watson J. (2016) Biodiversity: the ravages of guns, nets and bulldozers. Nature 536, 143–145.Munro N. and Lindenmayer D. B. (2011) Planting for Wildlife: A Practical Guide to Restoring Native Woodlands. CSIRO Publishing, Victoria, Australia.Pebesma E. (2018) Simple features for R: standardized support for spatial vector data. The R Journal 10 (1), 439–446. https://doi.org/10.32614/RJ‐2018‐009.R Core Team. (2020) R: A language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria. https://www.R‐project.org.RStudio Team (2020) RStudio: Integrated Development Environment for R. RStudio, PBC, Boston, USA. http://www.rstudio.com/.Shelley N., Shelley L., Purnell C. and Peter J. (2022) Wetland Birds App. BirdLife Australia, Victoria, Australia. [Accessed 13 Feb 2022.] Available from:. https://birdlife.org.au/projects/melbourne‐water‐biodiversity‐surveys/wetland‐birds‐app.Sievert C. (2020) Interactive Web‐Based Data Visualization with R, Plotly, and Shiny. Chapman and Hall/CRC, Boca Raton, Florida, USA. https://plotly‐r.com.Tobler M. W., Kéry M., Hui F. K. C., Guillera‐Arroita G., Knaus P. and Sattler T. (2019) Joint species distribution models with species correlations and imperfect detection. Ecology 100 (8), 2754. https://doi.org/10.1002/ecy.2754.Wickham H. (2016) ggplot2: Elegant Graphics for Data Analysis. Springer‐Verlag, New York. https://ggplot2.tidyverse.org.Williams J. and Price R. J. (2011) Impacts of red meat production on biodiversity in Australia: a review and comparison with alternative protein production industries. Animal Production Science 50, 723–747.Xie Y., Dervieux C. and Riederer E. (2020) R Markdown Cookbook. Chapman and Hall/CRC, Boca Raton, Florida. https://bookdown.org/yihui/rmarkdown‐cookbook.AppendixSpecies in BirdCastCommon nameScientific nameCommon BronzewingPhaps chalcopteraCrested PigeonOcyphaps lophotesPeaceful DoveGeopelia striataRainbow Bee‐eaterMerops ornatusSacred KingfisherTodiramphus sanctusLaughing KookaburraDacelo novaeguineaeGalahEolophus roseicapillusSulphur‐crested CockatooCacatua galeritaSuperb ParrotPolytelis swainsoniiRed‐rumped ParrotPsephotus haematonotusCrimson RosellaPlatycercus elegansEastern RosellaPlatycercus eximiusWhite‐throated TreecreeperCormobates leucophaeaBrown TreecreeperClimacteris picumnusSuperb Fairy‐wrenMalurus cyaneusNoisy FriarbirdPhilemon corniculatusLittle FriarbirdPhilemon citreogularisBrown‐headed HoneyeaterMelithreptus brevirostrisRed WattlebirdAnthochaera carunculataWhite‐plumed HoneyeaterLichenostomus penicillatusYellow‐faced HoneyeaterLichenostomus chrysopsSpotted PardalotePardalotus punctatusStriated PardalotePardalotus striatusWhite‐throated GerygoneGerygone olivaceaWestern GerygoneGerygone fuscaWeebillSmicrornis brevirostrisWhite‐browed ScrubwrenSericornis frontalisYellow‐rumped ThornbillAcanthiza chrysorrhoaYellow ThornbillAcanthiza nanaStriated ThornbillAcanthiza lineataBrown ThornbillAcanthiza pusillaBuff‐rumped ThornbillAcanthiza reguloidesGrey‐crowned BabblerPomatostomus temporalisBlack‐faced Cuckoo‐shrikeCoracina novaehollandiaeWhite‐winged TrillerLalage tricolorRufous WhistlerPachycephala rufiventrisGrey Shrike‐thrushColluricincla harmonicaCrested Shrike‐titFalcunculus frontatusPied CurrawongStrepera graculinaAustralian MagpieCracticus tibicenPied ButcherbirdCracticus nigrogularisGrey ButcherbirdCracticus torquatusWhite‐browed WoodswallowArtamus superciliosusDusky WoodswallowArtamus cyanopterusWillie WagtailRhipidura leucophrysGrey FantailRhipidura fuliginosaAustralian RavenCorvus coronoidesLeaden FlycatcherMyiagra rubeculaRestless FlycatcherMyiagra inquietaMagpie‐larkGrallina cyanoleucaWhite‐winged ChoughCorcorax melanorhamphosJacky WinterMicroeca fascinansMistletoebirdDicaeum hirundinaceumDiamond FiretailStagonopleura guttataBrown SonglarkCincloramphus cruralisRufous SonglarkCincloramphus mathewsiTree MartinPetrochelidon nigricansWelcome SwallowHirundo neoxenaSilvereyeZosterops lateralisCommon StarlingSturnus vulgarisTwo species that were in the large statistical model have not been included in BirdCast. The Australian Wood Duck (Chenonetta jubata) was not included due to its dependence on water. The Australasian Pipit (Anthus novaeseelandiae) is common in farm paddocks and relatively rarely observed inside woodland patches; it was excluded to avoid confusion. http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png Ecological Management & Restoration Wiley

A bird occupancy estimator for land practitioners in the NSW South Western Slopes bioregion

Download PDF
10 pages

Loading next page...
 
/lp/wiley/a-bird-occupancy-estimator-for-land-practitioners-in-the-nsw-south-yNga0yflFn

References (22)

Publisher
Wiley
Copyright
Copyright © 2022 Ecological Society of Australia and John Wiley & Sons Australia, Ltd
ISSN
1442-7001
eISSN
1442-8903
DOI
10.1111/emr.12556
Publisher site
See Article on Publisher Site

Abstract

IntroductionBiodiversity loss is a major global issue (Ceballos et al. 2017; Intergovernmental Science‐policy Platform on Biodiversity and Ecosystem Services (IPBES) 2019), and a significant amount of the past biodiversity loss has occurred in agricultural landscapes, both internationally (Maxwell et al. 2016; Filazzola et al. 2020) and in Australia (Williams and Price 2011). Globally, there have been major restoration efforts to recover biodiversity in agricultural landscapes (Brancalion & Holl 2020; Chazdon et al. 2020; Crouzeilles et al. 2020). However, greater information on the biodiversity dividends from restoration actions is often desired, including by on‐the‐ground practitioners responsible for implementing restoration programmes (Munro and Lindenmayer 2011). Such information is particularly important in extensively modified ecosystems such as the endangered Box Gum Grassy Woodland ecosystems, which characterise much of the wheat–sheep belt in northern Victoria, inland New South Wales and south‐east Queensland (Hobbs and Yates 2000). In these ecosystems, there have been long‐held concerns about declines in temperate woodland birds (Lindenmayer et al. 2018c; Ford et al. 2001). There also has been a need for information on the effectiveness of interventions such as plantings to reverse these declines (Belder et al. 2018; Lindenmayer et al. 2018b).We have developed a new, free webtool, BirdCast [https://sustfarm.shinyapps.io/BirdCast/], that has potential to describe the possible bird biodiversity dividends generated by investments in vegetation management and restoration programmes. BirdCast estimates the response of bird biodiversity in Box Gum Grassy Woodland to both existing woody vegetation and possible future woody vegetation after revegetation or natural regeneration in the NSW South Western Slopes bioregion (which also extends into northern Victoria) (DAWE 2012). We anticipate that BirdCast will be used by interested farmers to inform them about their farm's potential biodiversity and provide greater context to environmental management decisions. We anticipate that other natural resource management professionals will use BirdCast when engaging with farmers on management topics or decision‐making. BirdCast can produce estimates of:The chance of occupancy (occupancy probability) in one or more user‐specified woodland areas by 60 different bird species, including five species of conservation concern: Brown Treecreeper, Diamond Firetail, Dusky Woodswallow, Grey‐crowned Babbler and Superb Parrot (Department of Environment, Energy and Science 2021).Avian species richness.Bird biodiversity dividends from some management interventions, including:Addition or removal of nearby woody vegetation cover (i.e. vegetation restoration, natural regeneration or land clearing).Addition of new areas of planted eucalypt woodland.Interventions that prevent Noisy Miner occupancy of a woodland area.The webtool also builds user‐friendly key resources. These include the following:Reports for printing.Downloadable tables of estimates.Visualisations of estimates.A compilation of engaging bird images and descriptions (from BirdLife Australia and BirdLife Australia photographers) for the birds most likely to inhabit Box Gum Grassy Woodlands in farming landscapes.BirdCast is underpinned by robust statistical relationships developed from analyses of 17 years of bird observations (e.g. see Lindenmayer et al. 2018c, 2020) from 518 sites in temperate woodlands, primarily in New South Wales. BirdCast's quantitative estimates for scenarios at the scale of individual or multiple woodland areas make it a unique resource. Other resources typically provide species descriptions, species distribution maps or information on previous observations (Campbell et al. 2015; Atlas of Living Australia 2022; Shelley et al. 2022). These resources either do not have the quantitative depth of BirdCast or do not easily allow comparisons of biodiversity response to different management scenarios.This paper is an introduction to BirdCast for restoration practitioners and natural resource managers. We provide some background to the models and data used to create BirdCast (A Joint‐Species Statistical Model). The webtool was developed from the statistical model through user consultations and expert digital designers (Development of the Webtool). We then provide an example (A Worked Example), guidance on the interpretation of BirdCast outputs (Interpreting the Outputs of BirdCast) and limitations (Limitations).A Joint‐Species Statistical ModelThe first stage in the development of BirdCast (Fig. 1, left) was a statistical model for the occupancy of bird species in individual woodland areas. We used 5189 expert bird surveys in 518 different areas of remnant Box Gum Grassy Woodland or (re)planted woodland to create this model. These woodland areas were located primarily in the NSW South Western Slopes bioregion (DAWE 2012), with a lesser number of areas at locations up to the Queensland–NSW border (Fig. 2). The woodland areas were typically between 2 ha and 10 ha in area. For each woodland area, a representative survey site, 200 m long and 100 m wide, was selected. There was no minimum distance between different woodland areas, and more than 60% of sites were within 1 km of another site. The surveys were conducted in spring, spanning a time frame of 17 years. Surveys were generally repeated at least twice in a season to account for detection in the statistical model (Cunningham et al. 1999; Field et al. 2002). Planted woodlands were dominated by eucalypts 3+ years old, fenced at the time of planting and established using tubestock or direct seeding.1FigureDevelopment of BirdCast. Left: development of the statistical model. Right: development of the webtool.2FigureLocations of survey sites. Plus symbol (+): sites in remnant Box Gum Grassy Woodland. Green dots: sites in compatible planted woodland. Shaded region: NSW South Western Slopes bioregion (DAWE 2012). [Colour figure can be viewed at wileyonlinelibrary.com]The model included 62 species that were detected in 100 or more bird surveys. These species were mostly sedentary and land‐based. Five of the species were of conservation concern: Brown Treecreeper, Diamond Firetail, Dusky Woodswallow, Grey‐crowned Babbler and Superb Parrot (Department of Environment, Energy and Science 2021). The model accounted for interspecies correlation and imperfect detection (Tobler et al. 2019). Our method for selecting environmental predictors of species occupancy followed Hingee et al. (2022) and is the topic of the Supporting Information (Appendix S1–S3). The model passed a number of diagnostic tests (Section 5.5 of Appendix S1: Model fitting and selection), suggesting that estimates of occupancy probability, and estimated uncertainty of these estimates, can be trusted when the modelling assumptions are satisfied. We saw no indication of spatial correlation in the model's residuals for occupancy, despite the model having no spatially explicit dependence component. From our diagnostics, we suspect that much of the spatial correlation was taken on by the model's latent variables.Major site‐scale predictors of species occupancy were the presence of the hyper‐aggressive native honeyeater, the Noisy Miner (Manorina melanocephala), and whether the woodland area was planted versus remnant. Important covariates related to the surrounding landscape were the amount of woody vegetation cover within 500 m of the site centre and further away (within 3 km of the site centre), both of which were derived from satellite information (Liao et al. 2020). These woody vegetation cover data, sometimes termed foliage cover, were estimates of the proportion of land covered by 2 m + high branches or foliage (Liao et al. 2020). Australia's temperate woodlands typically have between 10% and 70% woody vegetation cover, depending on the age and structure of the vegetation (Hnatiuk et al. 2010). The woody vegetation cover estimates derived from satellite data may also fluctuate with water availability (Liao et al. 2020). Maps of the amount of woody vegetation cover can be viewed at [http://anuwald.science/tree], and a tool for obtaining the woody cover percentages within 500 m and 3 km of any location was integrated into BirdCast. Climatological averages were also crucial predictors of bird occupancy.Development of the WebtoolUsing R shiny (R Core Team, 2020; Chang et al. 2021), we built an initial webtool. In this version, important predictors and model outputs were shown on the same page. The user was required to input woody cover percentages limited to between 2% and 20%, which was approximately 90% of the range of woody cover percentages used to build the statistical model. We removed two species that risked confusing users: the Australian Wood Duck relies on waterbodies, so we expect the model estimates to be misinformative; and interpretation of occupancy probability estimates for the Australasian Pipit is prone to confusion as Australasian Pipits are much more likely to inhabit open farm paddocks than woodlands. The remaining 60 species in the webtool are listed in the Appendix A. Users were able to compare estimates with a previously saved scenario through graphics showing the relative increase or decrease (i.e. ratio) in species occupancy probability.When the user specified multiple woodland areas, the occupancy probability was defined as the chance that the species occupied at least one of the specified woodland areas. This form of occupancy probability was estimated by assuming that a species occupying one woodland area would also occupy any woodland areas that were more favourable according to our statistical model. This could be considered as a lower bound on the true occupancy probability – weaker statistical dependence between woodland areas will lead to higher occupancy probability estimates. Mathematically, the estimate for each species was the maximum of the occupancy probability estimate for each individual woodland area. This maximum is simple to compute from the results of our statistical model and produces higher values than an average of the estimates.This initial version was demonstrated to landholders, Local Land Services staff and other Natural Resource Management professionals at workshops in Orange and Wagga Wagga. A third workshop, scheduled for Wangaratta, was postponed due to COVID‐19. Detailed feedback on the webtool was also sought through consultations with potential users.We then engaged expert digital designers for further feedback and to design a new version of the webtool based on all prior feedback. This design had an easy linear workflow for users, was more touchscreen‐friendly, made comparisons easier to use, had cleaner graphs and moved engaging BirdLife Australia Photography images into the user's main view. We also incorporated interactive maps (Cheng et al. 2021) so that users could load satellite‐based estimates of woody cover by clicking on the centre of their woodland area.BirdCast can be used by visiting [https://sustfarm.shinyapps.io/BirdCast/] with any modern Internet browser on a desktop, tablet or phone. A demonstration video of BirdCast is available in BirdCast's user guide, accessed at the top of the webtool. Users will find it helpful to have some experience in using modern online interfaces, including online maps and satellite imagery. A live Internet connection is required. The source code for BirdCast is available at [https://github.com/sustainablefarms/farm_biodiversity_app/] and can be used with RStudio (RStudio Team 2020) to run BirdCast offline without the BirdLife Australia Photographs. BirdCast uses R shiny (R Core Team, 2020; Chang et al. 2021) and several packages written in the statistical programming language R (Wickham 2016; Pebesma 2018; Sievert 2020; Xie et al. 2020; Cheng et al. 2021; Coene 2021).A Worked ExampleStep 1: Your farmSuppose we are interested in a remnant area of Box Gum Grassy Woodland in the Cowra region of NSW. Noisy Miners occupy this area, and it has 9.5% and 7.5% woody vegetation cover within 500 m and 3 km, respectively, of the area's centre (these values were the average values for our survey sites in the Cowra region during the years that bird surveys were conducted). We provide all this information in Step 1 of BirdCast to create Scenario 1 (Fig. 3). If we knew the GPS co‐ordinates for the location of the woodland area, then BirdCast could obtain the woody vegetation cover estimates from satellite data (Liao et al. 2020).3FigureIn Step 1 of BirdCast, the user provides their region (left) and attributes of their woodland areas (right). This information becomes Scenario 1 in BirdCast. [Colour figure can be viewed at wileyonlinelibrary.com]Step 2: Bird diversityIn Step 2, BirdCast estimates the occupancy probability of bird species for the woodland area (Fig. 4). We can see that, on average, approximately 12 of the 60 species in BirdCast are estimated to occupy the woodland area in Scenario 1. In comparison, the average woodland area in our data set is estimated to have approximately 14 species. If woody vegetation cover within 500 m of the centre of the Scenario 1 woodland area was equal to 2%, then BirdCast estimates that there would be approximately 1 fewer species occupying the woodland on average. If the woody vegetation cover within 500 m was equal to 20%, then BirdCast estimates that there would be approximately 2 additional species occupying the woodland area on average.4FigureIn Step 2, BirdCast presents occupancy probability estimates for Scenario 1. The top display shows estimates of expected species richness under Scenario 1, in an average woodland area in our data set, and under two variations to Scenario 1. Other displays show the most likely and least likely species according to the occupancy probability estimates, details for the five vulnerable species in BirdCast, and a figure of occupancy probability estimates for all species in BirdCast. [Colour figure can be viewed at wileyonlinelibrary.com]Step 2 also includes displays of the 10 most likely species and 10 least likely species to occupy the woodland area and the five vulnerable species in BirdCast. These displays include engaging images from BirdLife Australia Photography, with further information displayed when clicking on an image. A final display shows the occupancy probability estimate for every species in BirdCast, and a table of these estimates is available for download. A full pdf report of Step 2 can be downloaded at the bottom of the page.Steps 3 and 4: Create a new scenario and compare bird diversityUsing BirdCast, we can view the estimates of bird species likely to occupy woodland areas and compare them with the species we would expect to see given changes to some key habitat and landscape attributes – such as fewer nearby trees, a new planting and an absence of Noisy Miners (which represents an improvement in habitat condition).No Noisy MinersIn Step 3, we create a second scenario, Scenario 2. BirdCast helpfully prepopulates Scenario 2 from the information in Scenario 1. For Scenario 2, suppose Noisy Miners are absent. We select ‘no’ in answer to the question about Noisy Miner presence for the woodland area, and then save and close the woodland area window. Removing Noisy Miners from a woodland area, or removing their impact, may be possible by increasing midstorey (Lindenmayer et al. 2018a). We then move to Step 4, which compares bird diversity between Scenario 1 and Scenario 2. BirdCast estimates that removing Noisy Miners increases average species richness by approximately 1 species and that the 10 most likely species now include the Black‐faced Cuckoo‐shrike and Willie Wagtail (Fig. 5). From the vulnerable species display, we can see that BirdCast estimates that there is a 26% chance of the Brown Treecreeper occupying the woodland area in Scenario 2 (without Noisy Miners), and only a 10% chance in Scenario 1 (with Noisy Miners). From the relative occupancy probability display, we can see that BirdCast estimates that removing Noisy Miners improves the occupancy chances for nearly all smaller‐bodied species, especially the Brown Thornbill, Leaden Flycatcher and Silvereye.5FigureIn Step 4, BirdCast compares bird species occupancy estimates for Scenario 1 and Scenario 2. The displays are similar to the displays in Step 2. [Colour figure can be viewed at wileyonlinelibrary.com]Removal of woody vegetation coverWe go back to Step 3. Suppose Noisy Miners are present again and some trees on the edge of the woodland area are removed so that we reduce woody vegetation cover within 500 m of the woodland's centre to 7.5% (from an original value of 9.5%). We then move forward again to Step 4 of BirdCast to compare Scenario 1 with this new Scenario 2. BirdCast estimates that the change in woody vegetation cover between Scenario 1 and Scenario 2 slightly lowers the average species richness, although the average richness remains approximately 12 species. In the relative occupancy section, we can see that most species are less likely to occupy the reduced woodland area (Scenario 2), especially small‐bodied bird species, although a few species are more likely to occupy the woodland area, such as the Brown Songlark and the Tree Martin, which are commonly associated with more open or cleared landscapes.A new woodland areaIf a new 1 ha woodland area (e.g. a 100 m × 100 m block planting) is established close to the first woodland area, then this creates a new type of habitat (planted woodland) and increases the woody vegetation cover near the first woodland area. Based on typical woody vegetation cover fractions for remnant temperate woodlands, the new woodland area could eventually result in 0.4 ha of additional woody vegetation cover, which increases the woody vegetation cover near the first, remnant, woodland area to 10% (from an original of 9.5%). We assume Noisy Miners are present in both. In Step 3, for the existing (remnant) woodland area we set Noisy Miners as present and 10% nearby woody vegetation cover. We then add a woodland area, select ‘planted woodland’ and set it to have 10% woody vegetation cover with 500 m. In Step 4, BirdCast suggests that the additional woodland area increases the expected richness of the farm by approximately 3 species. Furthermore, BirdCast estimates that Yellow Thornbills, Rufous Whistlers, and Red Wattlebirds are each more than three times as likely to occupy at least one of the woodland areas in Scenario 2 compared with the single woodland area in Scenario 1.Interpreting the Outputs of BirdCastThe outputs of BirdCast are most reliable for woodland areas that are:Eitherremnant Box Gum Grassy Woodlands, or3+ years old eucalypt‐dominated plantings that were established by tubestock or direct seeding and fenced at the time of planting.In or near the South Western Slopes bioregion (DAWE 2012), roughly the lower inland slopes of the Great Dividing Range, from Benalla in Victoria to Dubbo in NSW.Within a modified agricultural landscape dominated by livestock grazing or mixed farming enterprises, that is a relatively cleared landscape.Additionally, the estimates of occupancy are for springtime and average climatic conditions, which characterised the climate of the South Western Slopes between 2000 and 2017.When multiple woodland areas are specified in BirdCast, then species occupancy probability estimates are for the occupancy of any of the areas. These woodland areas must be sufficiently close to each other that bird species can easily colonise more favourable patches. Furthermore, the estimates apply to the specified woodland areas, not the neighbouring paddocks.The relative occupancy probability for a given species is the ratio of two different estimates. The numerator is the estimate of occupancy probability for Scenario 2, and the denominator is the occupancy probability estimate for Scenario 1. The larger and more positive the ratio, the larger the improvement in occupancy probability over the reference scenario. For example, a relative occupancy probability of ‘3’ means that the species is estimated to be three times more likely to occupy the woodland area(s) in Scenario 2 than the woodland area(s) in Scenario 1.Accuracy of estimates for species occupancy probability, in the form of 95% highest posterior density intervals, can be shown as error bars in many of the displays in BirdCast. When the modelling assumptions are met (Section S4.1), there is a 95% chance that the true occupancy probability is within the error bars. This uncertainty is due to the uncertainty of the fitted values for the parameters in the statistical model. Inaccuracies induced by the maximum assumption for multiple woodlands (A Joint‐Species Statistical Model) are not included.The estimates of species richness in BirdCast should be viewed as estimates of the average number of common, land‐based bird species occupying the woodland area(s). This expected species richness was the sum of the occupancy probability estimates of all species in BirdCast. Due to the mathematical properties of statistical expectations, no threshold on occupancy probabilities was required. These expected species richness estimates are unlikely to be equal to actual avian species richness because, according to the statistical model for a single woodland area, the realised number of species will vary with a standard deviation of four from the estimated average. Corresponding error bars have not been included in BirdCast to reduce confusion and due to the computational burden of the species interaction components of the statistical model. Furthermore, the estimate ignores water birds and many rare bird species that were not included BirdCast (see the Appendix A for the list of bird species in BirdCast).LimitationsDue to the finite nature of our data, and the difficulty in jointly modelling bird species, there are many important situations that BirdCast cannot directly assist with. Due to the locations of our bird survey sites, BirdCast should not be used to estimate bird diversity outside the NSW South Western Slopes bioregion, within urban areas or in large nature reserves (e.g. national parks). Similarly, there are many rare and endangered species that are absent from BirdCast due to insufficient detection of these species. BirdCast does not directly estimate bird diversity given fencing, underplanting and different management of farm paddocks as there were insufficient data on these activities to include in our statistical model. Additionally, BirdCast estimates should not be used to estimate bird diversity after extreme events such as wildfires or long‐term climate change.When using BirdCast, it is also important to remember that the results do not account for:Interactions between species (apart from the impact of Noisy Miners on other bird species). Although the statistical model was able to incorporate some forms of interactions between species (Tobler et al. 2019), the computational burden was too high to include in BirdCast.Connectivity of landscapes, additional to the impact of nearby and regional woody vegetation cover – landscape connectivity may have little association with occupancy probability of bird species in this region (Lindenmayer et al. 2020).Presence of surface water (riparian waterways and dams).Randomness in the relationship between occupancy of nearby woodland areas. Our statistical model appears to have accounted for associations between nearby woodland areas through fitted latent variables (Section S5.5). However, these fitted latent variables do not easily allow for statistical prediction of these associations in unsurveyed woodland areas.Planned near‐term improvements to BirdCast include improved relative occupancy probability visualisations and faster loading of the interactive maps. There is potential for enabling estimates in regions where we have bird survey data (Fig. 1) further north of the NSW South Western Slopes bioregion. We are also particularly interested in improving BirdCast's estimates when multiple woodland areas are specified.ConclusionsBirdCast is a free webtool that provides farm‐specific bird occupancy estimates under different scenarios. It enables quantitative estimates of the bird species assemblages in response to some common restoration interventions and provides an opportunity to explore the effects of proposed interventions in an interactive model that can engage and motivate farmers. The tool may also be used as a prioritisation tool by natural resource management agencies to predict and direct restoration efforts to achieve the best bang for buck investment scenarios.AcknowledgementsWe acknowledge Dr Martin Westgate's crucial contributions to the development of BirdCast. Many valuable improvements were suggested by Suzannah Macbeth. Additional improvements were suggested by Colleen O'Malley, Michelle Young and early users of BirdCast. The authors thank the participants of our workshops and consultations, and the team at CRE8IVE. Quantities derived from spatial data played a crucial role in developing the model and were computed by Prof. Albert van Dijk. The authors acknowledge BirdLife Australia for the generous use of the species descriptions in Bird Cast and the following from BirdLife Photography for their photographs: Stephen Garth, Mark Lethlean, Michael Hamel‐Green, Glenn Pure, Jonathan Pridham, Edward Mikucki, Geoffrey Stapley, Lindsay Hansch, Woody Woodhouse, John Barkla, Kathy Zonnevylle, Richard Smart, Rob Parker, Trevor Bullock, Shane Baker, Ian Wilson, Chris Dubar, Anthony Thompson, Brian O'Leary, Bill Carroll and Con Boekel. Open access publishing facilitated by Australian National University, as part of the Wiley ‐ Australian National University agreement via the Council of Australian University Librarians.ReferencesAtlas of Living Australia (2022) Explore Your Area. CSIRO, Canberra, Australia. [Accessed 13 Feb 2022]. Available from URL:. https://biocache.ala.org.au/explore/your‐area.Belder D. J., Pierson J. C., Ikin K. and Lindenmayer D. B. (2018) Beyond pattern to process: current themes and future directions for the conservation of woodland birds through restoration plantings. Wildlife Research 45 (6), 473–489. https://doi.org/10.1071/WR17156Brancalion P. H. S. and Holl K. D. (2020) Guidance for successful tree planting initiatives. Journal of Applied Ecology 57 (12), 2349–2361. https://doi.org/10.1111/1365‐2664.13725.Campbell I., Woods S. and Leseberg N. (2015) Birds of Australia: A Photographic Guide. Princeton University Press, Princeton, New Jersey.Ceballos G., Ehrlich P. R. and Dirzo R. (2017) Biological annihilation via the ongoing sixth mass extinction signaled by vertebrate population losses and declines. Proceedings of the National Academy of Sciences of the United States of America 114 (30), E6089–E6096. https://doi.org/10.1073/pnas.1704949114.Chang, W., Cheng, J., Allaire, J. J. et al. (2021). Shiny: web application framework for R. https://CRAN.R‐project.org/package=shinyChazdon R. L., Lindenmayer D., Guariguata M. R., Crouzeilles R., Rey Benayas J. M. and Chavero E. L. (2020) Fostering natural forest regeneration on former agricultural land through economic and policy interventions. Environmental Research Letters 15, 43002. https://doi.org/10.1088/1748‐9326/ab79e6.Cheng, J., Karambelkar, B., & Xie, Y. (2021). Leaflet: create interactive web maps with the JavaScript “leaflet” library. https://CRAN.R‐project.org/package=leafletCoene, J. (2021). Waiter: loading screen for “shiny.” https://CRAN.R‐project.org/package=waiterCrouzeilles R., Beyer H. L., Monteiro L. M. et al. (2020) Achieving cost‐effective landscape‐scale forest restoration through targeted natural regeneration. Conservation Letters 13 (3), e12709. https://doi.org/10.1111/conl.12709.Cunningham R. B., Lindenmayer D. B., Nix H. A. and Lindenmayer B. D. (1999) Quantifying observer heterogeneity in bird counts. Australian Journal of Ecology 24, 270–277.(DAWE) Dept of Agriculture, Water and the Environment. (2012). Interim Biogeographic Regionalisation for Australia (IBRA), Version 7 (Subregions)—States and Territories. www.environment.gov.au/fed/catalog/search/resource/details.page?uuid={1273FBE2‐F266‐4F3F‐895D‐C1E45D77CAF5}Department of Environment, Energy and Science (2021) Species Found in the Grassy Woodlands Formation. New South Wales Department of Environment, Energy and Science, Australia. https://www.environment.nsw.gov.au/threatenedSpeciesApp/VegSpecies.aspx?vegName=Grassy+woodlands&habitat=F.Field S. A., Tyre A. J. and Possingham H. P. (2002) Estimating bird species richness: how should repeat surveys be organized in time? Austral Ecology 27, 624–629.Filazzola A., Brown C., Dettlaff M. A. et al. (2020) The effects of livestock grazing on biodiversity are multi‐trophic: a meta‐analysis. Ecology Letters 23, 1298–1309. https://doi.org/10.1111/ele.13527.Ford H. A., Barrett G. W., Saunders D. A. and Recher H. F. (2001) Why have birds in the woodlands of southern Australia declined? Biological Conservation 97 (1), 71–88. https://doi.org/10.1016/S0006‐3207(00)00101‐4.Hingee, K. L., Westgate, M. J. & Lindenmayer, D. B. (2022). Long‐Term Monitoring in Endangered Woodlands Shows Effects of Multi‐Scale Drivers on Bird Occupancy Journal of Biogeography, 49 (5). https://doi.org/10.1111/jbi.14353.Hnatiuk R. J., Thackway R. and Walker J. (2010) Vegetation. In: National Committee on Soil and Terrain, Australian Soil and Land Survey Field Handbook. CSIRO Publishing, Victoria, Australia. https://www.publish.csiro.au/book/5230.Hobbs R. J. and Yates C. J. (2000) Temperate Eucalypt Woodlands in Australia: Biology, Conservation, Management and Restoration. Surrey Beatty & Sons, NSW, Australia.Intergovernmental Science‐policy Platform on Biodiversity and Ecosystem Services (IPBES) (2019) IPBES Global Assessment Summary for Policymakers. United Nations.Liao Z., Van Dijk A. I. J. M., He B., Larraondo P. R. and Scarth P. F. (2020) Woody vegetation cover, height and biomass at 25‐m resolution across Australia derived from multiple site, airborne and satellite observations. International Journal of Applied Earth Observation and Geoinformation 93, 102209. https://doi.org/10.1016/j.jag.2020.102209.Lindenmayer D. B., Blanchard W., Crane M., Michael D. and Florance D. (2018a) Size or quality. What matters in vegetation restoration for bird biodiversity in endangered temperate woodlands? Austral Ecology 43 (7), 798–806. https://doi.org/10.1111/aec.12622.Lindenmayer D. B., Blanchard W., Crane M., Michael D. and Sato C. (2018b) Biodiversity benefits of vegetation restoration are undermined by livestock grazing. Restoration Ecology 26, 1157–1164.Lindenmayer D. B., Blanchard W., Foster C. N. et al. (2020) Habitat amount versus connectivity: an empirical study of bird responses. Biological Conservation 241, 108377. https://doi.org/10.1016/j.biocon.2019.108377.Lindenmayer D. B., Lane P., Westgate M. et al. (2018c) Tests of predictions associated with temporal changes in Australian bird populations. Biological Conservation 222, 212–221. https://doi.org/10.1016/j.biocon.2018.04.007.Maxwell S., Fuller R. A., Brooks T. and Watson J. (2016) Biodiversity: the ravages of guns, nets and bulldozers. Nature 536, 143–145.Munro N. and Lindenmayer D. B. (2011) Planting for Wildlife: A Practical Guide to Restoring Native Woodlands. CSIRO Publishing, Victoria, Australia.Pebesma E. (2018) Simple features for R: standardized support for spatial vector data. The R Journal 10 (1), 439–446. https://doi.org/10.32614/RJ‐2018‐009.R Core Team. (2020) R: A language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria. https://www.R‐project.org.RStudio Team (2020) RStudio: Integrated Development Environment for R. RStudio, PBC, Boston, USA. http://www.rstudio.com/.Shelley N., Shelley L., Purnell C. and Peter J. (2022) Wetland Birds App. BirdLife Australia, Victoria, Australia. [Accessed 13 Feb 2022.] Available from:. https://birdlife.org.au/projects/melbourne‐water‐biodiversity‐surveys/wetland‐birds‐app.Sievert C. (2020) Interactive Web‐Based Data Visualization with R, Plotly, and Shiny. Chapman and Hall/CRC, Boca Raton, Florida, USA. https://plotly‐r.com.Tobler M. W., Kéry M., Hui F. K. C., Guillera‐Arroita G., Knaus P. and Sattler T. (2019) Joint species distribution models with species correlations and imperfect detection. Ecology 100 (8), 2754. https://doi.org/10.1002/ecy.2754.Wickham H. (2016) ggplot2: Elegant Graphics for Data Analysis. Springer‐Verlag, New York. https://ggplot2.tidyverse.org.Williams J. and Price R. J. (2011) Impacts of red meat production on biodiversity in Australia: a review and comparison with alternative protein production industries. Animal Production Science 50, 723–747.Xie Y., Dervieux C. and Riederer E. (2020) R Markdown Cookbook. Chapman and Hall/CRC, Boca Raton, Florida. https://bookdown.org/yihui/rmarkdown‐cookbook.AppendixSpecies in BirdCastCommon nameScientific nameCommon BronzewingPhaps chalcopteraCrested PigeonOcyphaps lophotesPeaceful DoveGeopelia striataRainbow Bee‐eaterMerops ornatusSacred KingfisherTodiramphus sanctusLaughing KookaburraDacelo novaeguineaeGalahEolophus roseicapillusSulphur‐crested CockatooCacatua galeritaSuperb ParrotPolytelis swainsoniiRed‐rumped ParrotPsephotus haematonotusCrimson RosellaPlatycercus elegansEastern RosellaPlatycercus eximiusWhite‐throated TreecreeperCormobates leucophaeaBrown TreecreeperClimacteris picumnusSuperb Fairy‐wrenMalurus cyaneusNoisy FriarbirdPhilemon corniculatusLittle FriarbirdPhilemon citreogularisBrown‐headed HoneyeaterMelithreptus brevirostrisRed WattlebirdAnthochaera carunculataWhite‐plumed HoneyeaterLichenostomus penicillatusYellow‐faced HoneyeaterLichenostomus chrysopsSpotted PardalotePardalotus punctatusStriated PardalotePardalotus striatusWhite‐throated GerygoneGerygone olivaceaWestern GerygoneGerygone fuscaWeebillSmicrornis brevirostrisWhite‐browed ScrubwrenSericornis frontalisYellow‐rumped ThornbillAcanthiza chrysorrhoaYellow ThornbillAcanthiza nanaStriated ThornbillAcanthiza lineataBrown ThornbillAcanthiza pusillaBuff‐rumped ThornbillAcanthiza reguloidesGrey‐crowned BabblerPomatostomus temporalisBlack‐faced Cuckoo‐shrikeCoracina novaehollandiaeWhite‐winged TrillerLalage tricolorRufous WhistlerPachycephala rufiventrisGrey Shrike‐thrushColluricincla harmonicaCrested Shrike‐titFalcunculus frontatusPied CurrawongStrepera graculinaAustralian MagpieCracticus tibicenPied ButcherbirdCracticus nigrogularisGrey ButcherbirdCracticus torquatusWhite‐browed WoodswallowArtamus superciliosusDusky WoodswallowArtamus cyanopterusWillie WagtailRhipidura leucophrysGrey FantailRhipidura fuliginosaAustralian RavenCorvus coronoidesLeaden FlycatcherMyiagra rubeculaRestless FlycatcherMyiagra inquietaMagpie‐larkGrallina cyanoleucaWhite‐winged ChoughCorcorax melanorhamphosJacky WinterMicroeca fascinansMistletoebirdDicaeum hirundinaceumDiamond FiretailStagonopleura guttataBrown SonglarkCincloramphus cruralisRufous SonglarkCincloramphus mathewsiTree MartinPetrochelidon nigricansWelcome SwallowHirundo neoxenaSilvereyeZosterops lateralisCommon StarlingSturnus vulgarisTwo species that were in the large statistical model have not been included in BirdCast. The Australian Wood Duck (Chenonetta jubata) was not included due to its dependence on water. The Australasian Pipit (Anthus novaeseelandiae) is common in farm paddocks and relatively rarely observed inside woodland patches; it was excluded to avoid confusion.

Journal

Ecological Management & Restoration – Wiley

Published: May 1, 2022

Keywords: Avian; biodiversity estimates; Noisy Miner; planting; temperate woodlands; webtool

There are no references for this article.