Long-term changes in populations of rainforest birds in the Australia Wet Tropics bioregion: A climate-driven biodiversity emergency

>850). These elevational bands were selected to have the finest scale division of elevation pos- sible with a relatively equal band width and sample size in each band. Species were categorised as rainforest specialist if rainforest represented their main or core breeding habitat [52]. Eleva- tion specialisation was defined by the elevational abundance profile of each species using a mean abundance for all surveys conducted between 1996–2004 within each elevational band as the baseline elevation abundance profile (supplementary material). Species were assigned as a specialist in that elevational band when >70% of their elevational abundance profile occurred within that band. Population indices and trends Overall population indices and trends for all 42 species were modelled using the rtrim package [77], an R-package based on the TRIM software (Trends and Indices for Monitoring data. TRIM v. 3.54. [78]). TRIM is designed to analyse time-series of counts and produce unbiased yearly indices and standard errors using log-linear models. The programme also estimates the dispersion factor, correcting for over-dispersion, and takes account of serial correlation between counts at the same site in different years [78]. This method has proved to be robust in trend calculation with missing years [79–82]. We used model 2 in rtrim, which assumed all years as possible changing points in the population trend [83]. Overall trends for each species were calculated with both unweighted data and data weighted by the geographic area within

each of the elevational bands. The weighted trends give an estimate of the changes in the spe- cies total population size as it takes into account the area within each elevational band as well as local abundance changes. Given that the difference in results using weighted and unweighted data was very marginal (supplementary material) and our primary focus here was to examine site-specific changes in local abundance and elevational distribution shifts in abun- dance, we have only presented in the results section the unweighted trends. Area-weighted trends of total population size are included in the Supplementary material (https://doi.org/10. 25903/3tmd-sw18). Individual species indices produced by TRIM were combined into multi-species indicators for each predefined group. Multi-species indicators were calculated using the MSI-tool in R [84]. This tool uses species-specific annual indices and their standard errors provided by TRIM to calculate annual multispecies indicator with confidence interval, accounting for sam- pling error, using the Monte Carlo simulation method. This method calculates a mean and a SE from 1000 simulated multi-species indicators and back-transforms these to an index scale, then repeats the process 10000 times. Those indicators are considered a measure of biodiver- sity change, where a reduction in index mean will occur if more species are declining than increasing and vice versa [80,85]. We tested for significant differences between the multi-spe- cies indicators using the TREND_DIFF-function, based on a Monte Carlo procedure (1000

iterations) and report the average difference with SE in the multiplicative trends and the signif- icance of this difference. Additionally, we explored the influence of changing trends over the study period by sepa- rately examining the trends prior to, and after, 2008 (the midpoint of the time-series). This enables some consideration of how the trends would have been observed over shorter time periods. Finally, to examine the population changes of each species within each elevational band, we estimated the local trends for each species along the elevational gradient. Multi-species trends in each elevation category were combined for each of the elevation specialist species groups. Differences within groups were tested using the TREND_DIFF-function. R (version 3.6.2) was used in all the analyses [86]. PLOS ONE Rainforest birds on escalator to extinction PLOS ONE | https://doi.org/10.1371/journal.pone.0254307 December 22, 2021 11 / 16