Cavity-nesting bird populations are most frequently limited by the number of tree cavities available in second-growth forests. However, this possible limitation of a key resource is less clear in old-growth forests. We compared forest attributes (i.e., basal area, density of larger trees, density of dead trees, and tree cavity density) in second-growth and old-growth stands in Andean temperate rainforests in southern Chile. To examine the role of nest-site availability in limiting the populations of Thorn-Tailed Rayaditos (Aphrastura spinicauda), a secondary cavity-nesting bird species, we conducted an experiment in which nest-boxes were added and removed in old-growth and second-growth forests during a five-year period (2008-2013). In old-growth forests, as compared to second-growth forests, we found a more than double basal area (99.6 vs. 43.7 m2/ha), a three times higher density of larger trees (88.2 vs. 36.4 trees/ha), and a 1.5 times higher number of small cavities (25.9 vs. 10.3 cavities/ha). The density of cavities also strongly increased with tree diameter and basal area. In second-growth forests, Thorn-Tailed Rayaditos showed a strong response to the addition and removal of nest-boxes, with population abundance increasing by 13% and then decreasing by 50%, respectively. In contrast, we found no impact in old-growth stands. Our experiment emphasizes the importance of maintaining large and dead trees in second-growth, disturbed, and managed forests. These trees provide suitable cavities for Thorn-Tailed Rayaditos, and likely many other secondary cavity nesters, increasing their abundances in a Globally significant Biodiversity Hotspot in southern South America.

Study area

We studied a small bodied secondary cavity-nesting bird species in Andean temperate rainforests of southern Chile (Thorn-Tailed Rayadito, 11.74 g; Altamirano et al. 2020), in the La Araucanía Region (39º16’ S, 71º48’ W) (Fig. 1). The South American temperate rainforests are one out of only seven of these temperate rainforest ecosystems in the world (Alaback, 1991). These forests cover more than 40,000 km² along the south-western edge of the continent, mainly in Chile and a small adjacent area in Argentina (CONAF et al., 1999; Donoso, 1993), stretching in latitude south between 35° and 55° (Armesto et al., 1998). They are characterised by cool summers, with precipitation occurring throughout the year. Furthermore, they have been classified as among the world’s 200 biologically most valuable and critically endangered ecoregions (Olson and Dinerstein, 1998) and are considered a Global Biodiversity Hotspot (Myers et al., 2000). In Andean areas, native forest has been reduced by 44% over the last two decades (Altamirano and Lara, 2010), and in coastal areas, 67% of native forests have been replaced by other types of land cover since 1975 (Echeverria et al., 2006).

We quantified forest attributes, cavity density, and the abundance of a secondary cavity nester (SCN) in 10 forest stands with different degrees of disturbance in the Andean Cordillera (Fig. 1). The stands ranged from recently burned areas with very low and occasional selective logging, representing second-growth forests (< 80 years), to non-burned old-growth forests (> 200 years old; Caviedes and Ibarra, 2017). Six of the 10 stands corresponded to second-growth forests, mainly with a predominance of broadleaf species such as Nothofagus obliqua, Nothofagus dombeyi and Laurelia sempervirens. The other four corresponded to old-growth stands of conifer-broadleaf mixed forest, principally with Saxegothaea conspicua, Laureliopsis philippiana and N. dombeyi. Forest stands were considered as old-growth when they had a core area with a minimal edge effect and maintained a complex vertical structure and the mostly unmodified species composition (Armesto et al., 2009). In both second-growth and old-growth forest stands, understory composition was dominated by bamboo species (Chusquea spp.), Rhaphithamnus spinosus, and a variety of species of Azara and tree saplings. As the home range of the focal bird species is unknown, we assumed a range of 3-4 ha for the Thorn-Tailed Rayadito, given the observed home-range for its close relative, Aphrastura masafuerae (Hahn et al., 2010). Our study stands were separated by a minimum linear distance of 1.6 km, 3.3 to 3.8 times the median natal dispersal reported by Botero-Delgadillo et al. (2017; 490 m females and 420 m males) to ensure that populations in one stand would not affect birds inhabiting adjacent study stands (Wiebe, 2011).

Forest attributes and cavity density 

To quantify forest structural attributes and cavity density in the stands, we established five vegetation plots per stand. These plots had their centre at each point where the abundance of rayaditos was surveyed (details below), with a survey area of 0.04 ha (radius = 11.2 m). Within these vegetation plots, we quantified the following stand attributes: density of live trees, density of dead standing trees, and for all live and dead trees with a tree diameter at breast height (DBH) ≥ 12.5 cm, we measured DBH and the number of small cavities per tree. We used tree DBH to calculate the basal area of trees with DBH ≥ 35 cm and the density of larger trees (DBH ≥ 60 cm). Tree cavities, including round or square entrances, crevices and branch holes, among others, were considered if they had a minimum entrance diameter of 2.5 cm, and a maximum diameter of 5 cm (Cornelius, 2008; Altamirano et. al. 2017; Ibarra et. al. 2020). We counted cavities with a minimum depth of 10 cm from the lower lip of the entrance (Altamirano et al. unpublished data). Cavity heights between 0.2 m and 30 m on the trees were considered, consistent with reports from previous studies (Altamirano et al., 2012; Cornelius, 2008; McGehee et al., 2010).

Resource addition and removal

We increased the density of suitable cavities by installing 40 wooden nest-boxes in each of six forests (160 boxes in four second-growth forests and 80 in two old-growth forests; n = 240) during the 2010 non-breeding season. The remaining four forest stands (two second- and two old-growth forests) were used as control stands where nest-boxes were not installed. Nest-boxes were hung from branches at a height of 1.5 m. The great majority of nest boxes were installed in the forest interior. When there was an open area nearby, nest-boxes were placed at least 15 m in from the forest edge. The direction in which the entrance to the nest-boxes faced was random. The nest-box entrance diameter and depth of the box were 3.1 cm and 17.1 cm, respectively (Altamirano et al., 2013; 2015). Nest-boxes of this type (i.e., entrance diameter and internal dimensions) were used to improve the probability of their occupancy (Lambrechts et al., 2010) as they had previously been successfully used by Thorn-Tailed Rayaditos (Moreno et al., 2005; Vergara, 2007). In the winter of 2012, we blocked all nest-boxes, reducing cavity density back to its original level in both second-growth and old-growth forests.

Secondary cavity-nesting bird abundance

We used an experimental design that considered monitoring densities of Thorn-Tailed Rayaditos before and after the cavity addition and removal (Aitken and Martin, 2012; Robles et al., 2012). We monitored the Thorn-Tailed Rayaditos populations in experimental and control stands over a five-year period: two years of pre-treatment (2008-2010); two years of treatment (nest-box installation, 2010-2012), one year of post-treatment (nest-box blocked, 2013). Five-point count stations per forest stand were established, with a diameter of 50 m and a minimum distance of 125 m between stations. All counts were conducted between October and February, during the four hours after sunrise (06.30-10.30) and had a duration of seven minutes during which we recorded all the bird species that were detected (heard and/or seen) within the radius. Point count stations were selected instead of transects as they are more efficient in forest conditions (Ralph et al., 1996).

Data analysis

We compared the following forest structural attributes of second-growth and old-growth forests: basal area of trees with DBH ≥ 35 cm (m²/ha; hereafter, basal area), the number of trees with DBH ≥ 60 cm per ha (hereafter, density of larger trees), the number of dead trees per ha (hereafter, density of dead trees), and the number of small-sized cavities per ha (hereafter, cavity density). We tested statistical differences using the t-test for pairs comparison, testing first for normality (Shapiro Wilk test; p > 0.05) and homoscedasticity (Levene’s test; p > 0.05). A critical tree diameter was established: i) DBH of 35 cm for the basal area because 80% of the nests of rayaditos were in trees with a higher DBH, and ii) DBH of 60 cm to calculate the density of larger trees, given that this is the mean diameter of the trees used by this species (Altamirano et al. 2017). We used two independent linear regressions to assess how the density of natural cavities was related to two different scales: tree DBH and basal area. For the number of cavities per tree, we used log10(number of cavities +1) transformation to achieve the normality assumption. We assessed cavity limitation for the Thorn-Tailed Rayadito, analysing its responses to experimental cavity addition and removal. Although we did not account for detectability, we think this might not have affected our main conclusions, as we compared abundances using consistent methods across forest stands and seasons. To determine the response to experimental increasing and reduction of cavity densities, we used Linear Mixed-Effect Models with a Poisson distribution, and the Akaike’s Information Criterion (AIC) approach to select the best fit models. Forest type, nest-box treatment (pre-treatment, during treatment, and post-treatment) and their interaction were included as fixed effects. Season and forest stand were included as random effects, allowing us to i) control for any inherent capacity of each season to have lower or higher abundances of rayaditos (Altamirano et al. 2020) and ii) include repeated measurements in the same forest stands (pre-treatment, during treatment, and post-treatment). Models with DAIC < 2.0 were considered the best-supported models (Burnham & Anderson 2002).