Birds associated with treefall gaps in a lowland forest in southwestern Brazilian Amazonia

Natural communities are dynamic systems in time and space. The spatial distribution of plants and animals tends to coincide with the availability of resources needed for the survival and reproduction of each species. Natural treefall gaps offer a number of resources that influence the distribution of birds within the forest. We compared the understory bird assemblages of natural treefall gaps (15 sampling points) with those found in the adjacent forest (15 points) in the Humaitá Forest Reserve in southwestern Brazilian Amazonia. We used mist-nets to sample birds and obtained 700 captures of 105 species. Species richness, number of individuals, and species composition were all similar between gaps and forest, although six species presented some degree of association with the gaps, and nine with the forest. Nectarivores preferred gaps significantly over forest, whereas insectivores and frugivores were distributed equally between gaps and forest. Vegetation height and density differed between gaps and forest, and influenced the distribution of bird species in the two environments. Fruit availability was positively correlated with frugivore abundance in gaps. Overall, 33.3% of the birds associated with the treefall gaps are from lower forest canopy, while the others are understory species. We showed that the use of natural gaps by birds in a fragmented landscape of the Amazon forest contributes to the environmental heterogeneity and succession dynamics following natural events of habitat disturbance.


INTRODUCTION
Biological communities in natural forests are formed by a mosaic of microhabitats and their associated organisms (Whittaker and Levin 1977). Environmental variability, resource availability, and the adaptive capacity of different organisms will determine the composition of local assemblages (Townsend et al. 2010). In natural communities, species assemblages tend to be similar in environments that have similar abiotic conditions, such as climate, soil, and water (Thompson and Sorenson 2000). Species will be favored distinctly by environmental conditions, including those created in novel environments (Sousa 1984), leading to the unequal spatial distribution of natural communities (Scheiner et al. 2000). Mosaics of environments and their associated biological communities are mediated by events such as landslides (Garwood et al. 1979), wildfires (Silva et al. 2015), and natural treefall in forests, coupled with the succession process (Thompson and Sorenson 2000).
In forest ecosystems, natural treefall, including events caused by wind, creates open spaces within the forest, known as treefall gaps (Baker et al. 2016). The gaps promote ecosystem dynamics, help shape the structure of natural communities and contribute to the heterogeneity in the composition of tropical forests (Brokaw 1985;Wunderle et al. 1987;Levey 1988;Hubbell et al. 1999;Busing and Brokaw 2002;Lima 2005;Schnitzer and Carson 2010;Maranho and Salimon 2015;Terborgh 2017).
Treefall gaps have higher sunlight incidence and higher density of plants with young leaves than adjacent forest, which attracts herbivorous arthropods (Richards and Coley 2007) and are referred to as nutrient hotspots due to their high rates of decomposition and mineralization (Scharenbroch and Bockheim 2008). In forests, many organisms are adapted specifically to the spectrum of conditions and resources provided by treefall gaps, resulting in high abundance and diversity, including butterflies (Pardonnet et al. 2013), spiders (Peres et al. 2014), snails (Alvarez and Willig 1993), bats (Crome and Richards 1988), amphibians (Strojny and Hunter 2010), small mammals (Beck et al. 2004), and birds (Schemske and Brokaw 1981;Levey 1990;Wunderle et al. 2005).
Ornithological surveys of forest fragments in southwestern Amazonia have revealed a high diversity of species, many of which are rare and/or endemic, with poorly-known ecological characteristics (Guilherme 2001(Guilherme , 2012(Guilherme , 2016Rasmussen et al. 2005;Mestre et al. 2010). Our objective was to understand the influence of treefall gaps on the distribution of birds in a southwestern Amazonian forest. Specifically, we tested the hypotheses that (i) there are more species in treefall gaps relative to adjacent continuous forest due to their heterogeneity, and species composition and abundance differ between gaps and forest; (ii) some species have an intrinsic preference for gaps or forest; (iii) the vegetation structure and food resources (flowers and fruits) influence species richness, composition, and abundance of birds, so that trophic guilds of birds differ between gaps and forest; and (iv) the bird species associated with gaps are found in other specific forest strata.

Study area
We conducted the study in the Humaitá Forest Reserve (HFR), a 2,000-ha forest fragment administered by Universidade Federal do Acre (UFAC), in Porto Acre, Acre state, southwestern Brazilian Amazonia (9º45'19''S, 67º40'18''W; Figure 1). HFR is covered by open terra firme rainforest, interspersed with palm stands and patches of Guadua bamboo, as well as alluvial forest, known locally as várzea (Acre 2010;IBGE 2012). The climate is humid tropical, with mean annual temperatures of 24-26 ºC (Alvares et al. 2013), and mean annual rainfall of approximately 1,900 mm. The rainy season extends from October to April, and the dry season from May to September (Duarte 2006). See Pedroza et al. (2020) for details on the local bird fauna.
We considered two forest habitats in the HFR ( Figure  2a): natural treefall gaps and undisturbed continuous forest (henceforth gap and forest). A gap is 'a vertical opening in the forest extending through the canopy to within 2 m of the forest floor' (Brokaw 1982). We calculated the area of the gaps based on their maximum length and width (Wunderle et al. 2005). We defined one sampling site randomly within each gap (Figure 2b, c) and a control site in the adjacent forest, located 50 m away from the nearest gap edge (sensu Levey We sampled birds with mist-nets 12 × 2.5 m and 36-mm mesh, from May to November 2018, between 05h30 and 17h30. At each site, we arranged two nets in either a straight line or in L or T configuration, to fit into the gap, with the same layout being used in the forest. We banded all the birds captured on one tarsus with a permanent metal ring engraved with a unique numerical code provided by the Brazilian government's Centro Nacional de Pesquisa para Conservação das Aves Silvestres (CEMAVE/ICMBio) to EG's project # 1099, authorized by SISBIO license # 23269-1. We sampled each pair of gap-forest sites simultaneously on two consecutive days during four field sessions of 30 days, with intervals of 20-34 days between sessions. Overall, we sampled each site on eight days. We adopted the bird species nomenclature of Gill and Donsker (2019).
We classified each species as frugivore, insectivore, nectarivore, omnivore, or piscivore (sensu Wilman et al. 2014). Regarding the forest stratum in which the species typically forages, each species captured in the gaps was classified as terrestrial, understory, mid-story, canopy or secondary-growth species, based on Henriques et al. (2003) and Schulenberg et al. (2010). Terrestrial birds forage primarily on the ground, understory birds at heights of up to 5 m in shrubs and small trees, mid-story birds above 5 m and below the canopy, and secondary-growth birds exploit regenerating habitats.
We estimated the structure of vegetation in the gap and forest sites using the method of Wunderle et al. (2005). We obtained vegetation height profiles from two parallel transects, one located on each side of the mist-net, every 1 m along each transect using a 3-m pole as a reference scale. We recorded the presence or absence of vegetation touching the pole within each height interval. The height intervals (in meters) were: 0-0.5, 0.51-1, 1.01-2, 2.01-3, 3.01-4, 4.01-6, 6. 01-8, 8.01-10, 10.01-12, 12.01-15, 15.01-20, and > 20. An estimate of percentage cover was based on these data for each height interval. We evaluated the density of the vegetation for all gap and forest site using a modified version of the method of Levey (1988). We demarcated an area of 48 m 2 between the transects, within which we counted the number of trunks, shrubs, branches, and stems with a diameter ≥ 7 mm. We determined the absolute density of the plants in the plots using the formula of Freitas and Magalhães (2012). We quantified plants bearing flowers and/or fruits within the entire gap area below 10 m height (sensu Levey 1988). We collected a sample of each flowering or fruiting plant and we prepared exsiccates deposited and identified in the Laboratory of Botany and Plant Ecology (LABEV) at UFAC. Plant species nomenclature followed Daly and Silveira (2008) and Medeiros et al. (2014).

Data analysis
Our sampling units were 15 gaps compared to 15 forest sites. To compare species richness between gap and forest, after confirming the normality of the data (Sokal and Rohlf 1995), we applied a paired t test. We estimated species richness in gaps and forest using the nonparametric Jackknife 1 procedure. We evaluated the difference in the number of individuals captured in gaps and forest using the Chi-square test. We ordered the species composition in the gaps and forest through non-metric multidimensional scaling (NMDS) for abundance data using Bray Curtis similarity index. We ran the NMDS ordination through the function 'metaMDS' in the vegan package (Oksanen et al. 2018) in the R environment. To test for differences in species composition between gap and forest sites we ran a global test using Permutational Multivariate Analysis of Variance (PERMANOVA) (Anderson 2001) using the function 'adonis', employing 1000 permutations.
To identify habitat specialists and associates we ran an indicator value (IndVal) method (Cáceres and Legendre 2009) through the function 'indval' in the indicspecies package. This analysis takes into account specificity and fidelity. A perfect indicator species must occur only in samples of a certain category (specificity) and also in all the category units (fidelity). We estimated the significance of this analysis with a Monte Carlo test with 10000 randomizations through the function 'multipatt'. We classified the species in two categories: (a) habitat specialists (gap or forest), based on a p value ≤ 0.05; and (b) habitat associates, as the species captured at a frequency greater than 40% in gaps or forest.
To evaluate the variation in height and density of vegetation between gaps and forest, we applied the nonparametric Wilcoxon paired test. To evaluate the association between the number of flowering or fruiting plants in the gaps and the abundance and richness of nectarivore and frugivore species, respectively, we applied the Spearman correlation coefficient (r). To evaluate the influence of the vegetation structure on the composition and abundance of the bird assemblages, we used parcial redundance analysis (pRDA, see Legendre and Legendre 1998). To test if the model is significant, we ran a global test using Analysis of Variance (ANOVA), employing 1000 permutations.
To evaluate the variation in the frequency of trophic guilds between gaps and forest, and the frequency of the species present in each guild found in gaps and/or forest, we applied Chi-square tests. We calculated the percentage of use of the different foraging strata in gaps or forest by birds pooled in each specialist or associate category. We calculated the percentage of species that used specific strata, including only the species for which at least six individuals were captured in the mist-nets. We applied the paired Wilcoxon and Chi-square tests to evaluate the variation in the frequency of the different categories. We ran all analyses in the R software, version 3.5.1 (R Core Team 2018).

RESULTS
Total sampling effort was 3,916 h. We captured 700 birds belonging to 105 species and 26 families (Supplementary Material, Table S1). We banded 531 individuals and recaptured 144. We captured a similar number of individuals in gaps (366) and forest (334), with no significant difference between habitats (χ 2 = 1.4, df = 1, p = 0.22). Species richness was similar between gaps and forest, 81 and 80 species, respectively (t = 0.66, df = 14, p = 0.52). The species rarefaction curve did not present a tendency to stabilize, with an estimated total of 101 species for the gaps and 103 for the forest. The composition of the bird assemblages did not vary significantly between gaps and forest (PERMANOVA: F = 1.40, df = 1, p = 0.09, stress = 0.25).
We detected an association with gaps or forest in 15 species (Table 1), corresponding to 14.2% of the 105 recorded species. We classified three (20%) of these species as habitat specialists, two of which were specialized in gaps, and one in forest (Table  1). Four species (26.6%) were associated with gaps and eight (50%) with forest ( Table 1). The capture frequency of species with some degree of habitat association was similar between gaps and forest (χ 2 = 1.81, df = 1, p = 0.17). The recapture data ( Figure 3) indicate that Thalurania furcata, Sciaphylax hemimelaena, Xenops minutus, Lepidothrix coronata, Pipra fasciicauda and Mionectes oleagineus moved either between gaps or between gap and forest (Table 2).
The height (W = 2, p = 0.001) and density (W = 108, p = 0.001) of the vegetation varied significantly between gaps and forest. We counted 3,218 plants in the gap plots (214.5 ±   Figure 5). Less than 1% of the plants in the gaps were over 20 m height. The gaps had a mean size of 238.1 m 2 (range: 104-437.4 m 2 ). The pRDA explained 72.5% of the total variation in bird species abundance in relation to vegetation structure. The bird assemblages were influenced by the vegetation structure (F = 1.75, df = 2, p = 0.001; adjusted R 2 = 0.112). The first pRDA axis indicated a negative correlation (-81.7%) of the variation in total abundance with the height of the vegetation and also indicated a negative correlation (-55.7%) between vegetation density and the occurrence of some bird species. The second axis indicated that 43.5% of this variation is explained by the density of the vegetation.
We recorded 19 plants bearing flowers or fruit, representing 13 genera and 12 botanical families (Supplementary Material, Table S2). Abundance (r = 0.06, p = 0.02) and richness (r = 0.49, p = 0.05) of frugivore birds and fruit supply in the gaps showed a moderate positive relationship. However, we found no relationship between the number of flowering plants and the abundance (r = -0.3, p = 0.25) and richness (r = -0.2, p = 0.31) of nectarivores.
The two gap-specialist birds were recorded in the forest understory (100%), and the only forest specialist in the subcanopy (Table 1). Species associated with gaps were recorded in almost all strata (Table 1), primarily in the understory (71.4%) and sub-canopy (57.1%). Forest-associated species were recorded primarily in the understory (75%), and subcanopy and forest floor both 37.5%.
Two species that we classified as gap-associated were recorded primarily in the forest sub-canopy: Sittasomus griseicapillus and T. schistogynus. Three (37.5%) of the eight species associated with the forest also were recorded preferentially in the sub-canopy: Dendrocincla fuliginosa, Glyphorhynchus spirurus, and I. hauxwelli.

Species richness and composition, and bird abundance
Bird species richness did not vary significantly between gap and forest because most species were captured in the two types of habitat, as recorded in forests of Puerto Rico and Costa Rica (Wunderle et al. 1987;Levey 1988). Thus we could not confirm our assumption that bird assemblages in gaps and forest were different, as reported in Panama and Malaysia (Schemske and Brokaw 1981;Rosely et al. 2007). In the HFR, the availability of food resources in gaps was probably similar to that of the forest, contributing to the similarity in the bird assemblages, in contrast with the pattern suggested by Blake and Hoppes (1986) and Martin and Karr (1986).

Habitat preferences
We identified species specialized in the exploitation of either treefall gaps or continuous forest and others that were associated with both environments, i.e., species that have some degree of affinity with one habitat, but are not dependent on this environment in the way specialists are. This indicates that a minimal degree of differentiation exists between gap and forest bird assemblages. However, as the gaps found within a given forest will typically present different stages of maturation (Brokaw 1985), we would expect to find an overlap between the bird assemblages in gaps and forest. We suggest that few  forest species have specific adaptations for the exploitation of gap environments at an early stage of regeneration. This would account for the fact that we recorded only two species that were associated strongly with the treefall gaps in the HFR.
In the HFR, we captured only a few individuals of Arremon taciturnus and Cyanocompsa rothschildii, which are associated clearly with Amazonian treefall gaps (Schulenberg et al. 2010). It seems likely that these species may have a preference for larger gaps than those found in the HFR, or occur locally at very low densities. Overall, most species classified as specialists in both habitats in the HFR preferred gap environments, as found in other tropical forests (Schemske and Brokaw 1981;Wunderle et al. 2005). The HFR and El Verde in Puerto Rico had relatively few species associated with gaps. In some tropical forests, species richness and total captures are much higher in gaps than in forest (Levey 1988;Wunderle et al. 2005).

Habitat specialist birds
We captured individuals of Thamnomanes ardesiacus only in the forest, probably because this species inhabits primarily the understory of terra firme forest (Zimmer and Isler 2003) and, in the HFR, it seems likely that it avoids gaps. Hypocnemis peruviana was associated with gaps in the HFR, a behavior well documented in the Peruvian Amazon (Schulenberg et al. 2010) and also observed for Hypocnemis cantator (currently Spix's Warbling-antbird Hypocnemis striata, Isler et al. 2007) in the Tapajós National Forest (Wunderle et al. 2005). By contrast, Oryzoborus angolensis normally inhabits open areas, natural grassland beyond the forest edge, and secondary forest (Rising and Jaramillo 2011). It is a gap specialist in the HFR. Gaps in the early stages of regeneration offer a range of resources (such as seeds) capable of maintaining this species within the forest, as is the case with its congeners Sporophila aurita (S. corvina) in Costa Rica (Levey 1988) and Sporophila lineola in Brazil (Banks-Leite and Cintra 2008), both known to colonize continuous forest by exploiting gaps efficiently.

Bird movement between gaps and forest
We showed the actual use of gaps by birds which were recaptured in different gaps in the HFR. This reveals that the birds are moving systematically to these environments to forage, rather than just ranging randomly. The movements of Pipra fasciicauda and Lepidothrix coronata between gaps indicate that they play an important role in dispersing seeds from the forest to gaps and also between gaps (Murray 1988;Snow 1981;Loiselle and Blake 1990;Marini 1992;Piratelli and Mello 2001;Oliveira and Dario 2018).
We observed that Sciaphylax hemimelaena and Xenops minutus, which are common in secondary growth (Laurance et al. 2004), frequently forage in forest edges or gaps in the HFR. Sciaphylax hemimelaena was observed almost daily in the edges of the study gaps (J.L. personal obs.). Probably edge insectivores, birds that exploit gaps, and some frugivores move within the forest via these environments (Laurance et al. 2004). This behavior appears to be advantageous for many bird species, especially where the forest has undergone significant fragmentation (Lees and Peres 2009). Our results indicate that these movements do occur, and that this pattern may be frequent among all species associated with gaps in the HFR.

Distribution of trophic guilds
The number of plants flowering and fruiting in gaps was low in the HFR throughout the study period. We observed the nectarivores Amazila lactea, Chlorostilbon mellisugus and Phaethornis ruber foraging constantly within the gaps (J.L. personal obs.), but they were rarely captured. This indicates that, while these species were present in the gaps, our capture method may have been inadequate for the effective sampling of the members of this guild. The low capture rates of frugivores and nectarivores were likely related to the relative scarcity of food resources in the gaps during the study period (Gentry and Emmons 1987), which coincided with the dry season in the study region, which lasts from May to October (Duarte 2007). The same pattern was recorded in Puerto Rico and central Amazonia (Wunderle et al. 1987;Henriques et al. 2003). Despite the low capture rates, the positive correlation we observed between fruit availability and frugivore abundance corroborates the hypothesis that resource availability attracts these birds to the gaps (Willson et al. 1982;Blake and Hoppes 1986;Gomes et al. 2011).
In the HFR, P. fasciicauda was the most abundant frugivore, and was the species most captured in gaps, especially at the sites with a higher concentration of fruiting plants of the families Rubiaceae and Olacaceae. Fruits of Siparuna, Heisteria, Cissus, Psychotria, and Costus are components of the diet of many Neotropical frugivorous birds (Snow 1981), and gaps function as a "key habitat" for the maintenance of populations of frugivorous birds during periods of fruit shortage (Levey 1990). Manakins specialize in eating small fruit (Snow 1981), in particular those of Melastomataceae and Rubiaceae (Piratelli and Mello 2001). Melastomataceae are pioneer species (Mentz and Oliveira 2004) that grow and fruit quickly, taking advantage of the greater input of sunlight in gaps in the forest (Uhl et al. 1988;Brokaw 1985). This may explain why P. fasciicauda was the most abundant frugivore in gaps in the HFR.
One gap specialist in the HFR was an insectivore. The high species richness and the abundance of individuals in comparison with other trophic guilds (Wunderle et al. 2005;Neto et al. 2017) corroborate the hypothesis that high concentrations of seedlings and young leaves in early regrowth likely attracts many invertebrates to the gaps (Richards and Coley 2007). Insectivores may be less sensitive to the marked microclimate variation between gap and forest (Wunderle et al. 2006). The even distribution of insectivores in gaps and forest in the HFR is can be related to the active foraging behavior ACTA AMAZONICA of these birds, which move constantly through the forest in search of prey, often in mixed-species flocks dominated by insectivores (Munn and Terborgh 1979), which are common in the HFR (Pedroza et al. 2020), in both gaps and forest (J.L. personal obs.). For example, Thamnomanes schistogynus brings together a number of other species that form high-density, mixed-species flocks that forage together (Munn 1985). This accounts for the capture of a large number of insectivores in the gaps when T. schistogynus was present.

Influence of vegetation structure and forest strata
In the HFR, species influenced by the vegetation structure of the gaps used the sites with denser vegetation and taller plants, i.e., they prefer gaps at an intermediate or advanced stage of regeneration (Banks-Leite and Cintra 2008). In this stage, the vegetation of gaps generally is relatively dense, offering more substrates for potential food resources such as invertebrates (Didham et al. 1996). The greater height of the vegetation in these gaps also creates different strata that are exploited by a greater variety of bird species, especially insectivores (Felton et al. 2008). Advanced regrowth in gaps thus allows some bird species that normally occupy the highest forest strata to exploit this newly-formed environment. Surprisingly, most of the species that were common in gaps inhabit the sub-canopy and canopy in forest (Schulenberg et al. 2010).
Tachyphonus luctuosus, Veniliornis affinis and Piculus leucolaemus, which all inhabit the forest sub-canopy and canopy, usually were observed in gaps foraging in mixedspecies flocks (J.L. personal obs.). Forest sub-canopy birds are more commonly found in gaps in comparison with canopy birds, given the greater proximity of this stratum to the gap vegetation (Walther 2002). The movement of species from higher forest strata to the understory reflects vertical mobility, as observed in birds in many different rainforests (Schemske and Brokaw 1981;Wunderle et al. 1987;Wunderle et al. 2005). Given this, the occurrence of different species at the level of the mist-nets will vary among forest types (Levey 1988). In the HFR, this reflects a real preference for the gap environment in sub-canopy species, as observed by Schemske and Brokaw (1981) in a tropical forest in Panama, rather than just a tendency to occur in this environment as a random consequence of their movement patterns in the higher strata.

CONCLUSIONS
The assemblage of understory birds found in natural gaps was similar in species composition and richness to that of the adjacent forest in southwestern Brazilian Amazonia. Yet we identified species that specialize in the use of natural gaps, and others that are closely associated with these environments, but are not dependent on them. Natural gaps are exploited by birds from different strata of the forest, ranging from the ground to the sub-canopy and canopy. The different stages of gap maturation attract an ample diversity of birds that are fundamental to the regeneration and restoration of the forest environment. We suggest that the association of birds with natural forest gaps establishes a link that likely ensures pollination, seed dispersal, and pest control, and guarantees the heterogeneity and resilience of the forest environment following local impacts.

AKNOWLEDGMENTS
To Coordenação de Aperfeiçoamento de Pessoal de Nível Superior (CAPES) for the scholarships that were essential for financing the expeditions. To Universidade Federal do Acre (UFAC) through its Graduate Program in Ecology and Management of Natural Resources (MECO), for permission and logistical support during the fieldwork in the HFR. To Jhonnatan Machado, for support in the field, which guaranteed access to the gaps. To the team of the UFAC Ornithology Laboratory, for their help in the laboratory and fieldwork in the HFR. To Salatiel Clemente, for help with the statistical analyses. To Marcos Silveira (LABEV) and Evandro Ferreira (INPA), for the identification of the plant specimens. To the reviewers, for important suggestions and criticism that improved an earlier version of this manuscript. Table S1. Number of birds per species captured per habitat (15 treefall-gap and 15 forest sites) in the Humaitá Forest Reserve, southwestern Brazilian Amazonia. Guild: F = frugivore, I = insectivore, N = nectarivore, O = omnivore, P = piscivore. Stratum: T = terrestrial, U = understory, Sc = sub-canopy, C = canopy, Sf = secondary forest, Fe = forest edge, Tf = treefall. *Treefall gap specialist in the HFR. † Forest specialist in the HFR. Nomenclature and taxonomic orders follow Gill and Donsker (2019). Guild classification follows Wilman et al. (2014). Foraging strata follow Henriques et al. (2003) and Schulenberg et al. (2010).

Family/Species
Gap