How are rainforest birds faring in the Anthropocene? We use bird captures spanning >35 years from 55 sites within a vast area of intact Amazonian rainforest to reveal reduced abundance of terrestrial and near-ground insectivores in the absence of deforestation, edge effects, or other direct anthropogenic landscape change. Because undisturbed forest includes far fewer terrestrial and near-ground insectivores than it did historically, today’s fragments and second growth are more impoverished than shown by comparisons with modern ‘control’ sites. Any goals for bird community recovery in Amazonian second growth should recognize that a modern bird community will inevitably differ from a baseline from >35 years ago. Abundance patterns driven by landscape change may be the most conspicuous manifestation of human activity, but biodiversity declines in undisturbed forest represent hidden losses, possibly driven by climate change, that may be pervasive in intact Amazonian forests and other systems considered to be undisturbed. 

These count data were collected using a standard operating procedure of avian mist-netting across time (1980-1984 and 2007-2016) and space (70 unique mist net lanes) at the Biological Dynamics of Forest Fragments Project, ~80 km north of Manaus, Brazil (2º20’S, 60ºW). For specifics about field protocols, data quality control, subsetting, and categorization, we refer the reader to the pertinent sections of the Methods: Site selection, Bird sampling, and Bird data.

For starters, we would direct anyone interested in these data to begin by examining Figure 1 in the main text. This provides an illustrative schematic of our sampling design, as well as the three combinations of time period+forest type that we compared: historical primary, modern primary, and modern disturbed. These three samples directly align with the three principal datasets that we uploaded separately to Dryad (e.g., Historical_primary_forest_captures_BDFFP.csv). For each of the three forest samples (differentiated in the 'Time_period' column), we provide the number of captures ('Count') for all 'Species' at a given 'Site', along with the 'Effort' (in net-hours) that accrued at each site. Even though some datasets completely lack captures for a given species (Count = 0), each file nonetheless contains a row for every combination of site and species (n=79 species). However, because the datasets differ in the number of sites sampled, they also differ in length:

Historical primary forest - 34 sites*79 species = 2686 rows

Modern primary forest - 21 sites*79 species = 1659 rows

Modern disturbed forest - 15 sites*79 species = 1185 rows

These count data were then used to make relative abundance comparisons between samples in order to detect long-term changes within primary forest (Figure 2A) as well as the shifting baseline (Figure 2B). We also used a categorical column for 'Guild' to more generally describe long-term changes within primary forest (Figure 3A) and following deforestation and fragmentation (Figure 3B).

Additionally, "NMDS_community_ordination_BDFFP.csv" integrates captures and effort to generate capture rates (captures/1000 mist-net hours) for all 79 species (now 79 columns) at each of our 70 sites (rows): 15 sites in modern disturbed, 34 in historical primary, and 21 in modern primary forest. We used these data to generate the NMDS ordination evident in Figure 4.

For more information, please see the Methods section of the main text and Table S1 for a list of all 79 species, with summed captures across sites for each of the three datasets. Finally, we provide the coordinates for each of the 70 sites in UTMs (Zone 21S): "BDFFP_site_coordinates_21S.csv".