Fire regimes shape biodiversity: responses of avian guilds to burned forests in Andean temperate ecosystems of southern Chile

Fire regimes of forests, i.e., time interval, frequency, extent, and severity of fire events, influence structural changes in the vegetation, and thus shape the composition of avian communities. We studied the diversity of avian guilds in sites with different fire regimes (unburned, burned 2002, burned 2015, and burned 2002 and 2015), testing both the “intermediate disturbance” and “vertical vegetation structure” hypotheses, in globally threatened temperate forests in Chile. From 2016 to 2018, we quantified habitat attributes (160 plots) and estimated avian richness and density (160 point counts). The site that was burned once in 2015 showed the highest density of standing dead trees at 96.5% higher than the unburned/control site, whereas the site that burned twice showed the lowest density of live trees, lowest average diameter at breast height of trees (DBH), and smallest volume of coarse woody debris. Overall, we recorded 35 avian species with the highest richness (n = 24 species) in the site that was burned once in 2002. We found that, 16 years after a site was burned, the avian community composition became relatively similar to the unburned site. The density of most avian guilds decreased in burned sites but granivores, shrub users, and migrants showed positive responses. Understory users, foliage users, and resident species showed negative responses to burned sites. These responses were strongly related to fire-driven changes in habitat attributes, supporting both of our tested hypotheses. Given that increasing levels of disturbance from fire are anticipated, future management of temperate forest biodiversity should consider that specific species and guilds will depend on remnant habitat attributes in burned sites. Le régime des incendies façonne la biodiversité : réactions des guildes aviaires aux forêts brûlées dans les écosystèmes andins tempérés du sud du Chili RESUME_. Le régime des incendies de forêt, c'est-à-dire l'intervalle de temps, la fréquence, l'étendue et la sévérité des incendies, influent sur les changements structurels de la végétation et donc sur la composition des communautés aviaires. Nous avons étudié la diversité des guildes aviaires dans des sites présentant différents régimes d'incendie (non brûlé, brûlé en 2002, brûlé en 2015, et brûlé en 2002 et 2015), en testant à la fois les hypothèses de « perturbation intermédiaire » et de « structure verticale de la végétation » dans des forêts tempérées du Chili menacées à l'échelle planétaire. De 2016 à 2018, nous avons mesuré les caractéristiques de l'habitat (160 parcelles) et calculé la richesse et la densité aviaire (160 dénombrements par points d'écoute). Le site qui a brûlé une fois en 2015 présentait la plus forte densité d'arbres morts sur pied, soit 96,5 % de plus que le site non brûlé/témoin, tandis que le site qui a brûlé deux fois présentait la plus faible densité d'arbres vivants, le plus faible diamètre moyen à hauteur de poitrine des arbres (DHP) et le plus petit volume de débris ligneux grossiers. Nous avons noté 35 espèces aviaires dans l'ensemble, et la plus grande richesse (n = 24 espèces) se trouvait dans le site qui a brûlé une fois en 2002. Nous avons constaté que, 16 ans après qu'un site ait brûlé, la composition de la communauté aviaire était redevenue relativement similaire à celle du site non brûlé. La densité de la plupart des guildes aviaires a diminué dans les sites brûlés, mais les granivores, les utilisateurs d'arbustes et les migrateurs ont eu des réactions positives. Les utilisateurs du sous-étage, les utilisateurs du feuillage et les espèces résidentes ont réagi négativement aux sites brûlés. Ces réactions étaient fortement liées aux changements des attributs de l'habitat induits par le feu, résultat qui soutient nos deux hypothèses. Étant donné qu'on prévoit une augmentation des perturbations par le feu, la gestion future de la biodiversité des forêts tempérées devrait tenir compte du fait que des espèces et des guildes spécifiques dépendront des attributs de l'habitat restant dans les sites brûlés.


INTRODUCTION
Fires are critical drivers of the dynamics of several forest ecosystems because they modify the vertical and horizontal stand structure, often with long-lasting effects on the habitat of forestdwelling species (Flannigan et al. 2009, Stephens et al. 2015, Fuentes-Ramirez et al. 2018. Fires can, therefore, be a major cause of habitat loss because 4% of the earth's land surface is burned annually (30-46 million km²; Randerson et al. 2012). Fire is a natural disturbance of ecosystem processes in a number of ecoregions, e.g., boreal forests and savannah ecosystems, whereas in others the frequency of natural fires is intrinsically low, e.g., tropical Southeast Asia (Bond et al. 2005). Fire regime refers to the dynamic behavior of fires occurring over an extended period of time (Brown 1995). These regimes are being altered in many ecosystems worldwide by anthropogenic causes, including climate change (Harvey et al. 2016, Prichard et al. 2017. The effects of fire regimes on biodiversity will be influenced by the number of times a site has been burned, the time elapsed since the last fire, the extent of the fire, and its severity, i.e., the quantity of loss of organic matter above and below ground (Keeley 2009, Krebs et al. 2010, Chavardès 2014. Thus, fire regimes shape the structural complexity of post-fire habitats and may model the evolutionary adaptations of wildlife species in fire-prone ecosystems (Watson et al. 2012, Drapeau et al. 2016).
According to the intermediate disturbance hypothesis, habitat specialist species will predominate numerically in the community in undisturbed habitats (Connell 1978). In highly disturbed ecosystems, habitat generalist species, able to resist or adapt to the new environmental conditions, will persist (DeBano et al. 1998, Jaksic andFariña 2015). In ecosystems with slight to moderate levels of disturbance, diversity will be maximized because both specialists and generalists may coexist (Connell 1978). Although some theoretical and empirical studies have rejected this hypothesis (e.g., Fox 2013), it still provides an important conceptual framework for examining the diversity found across habitat disturbance gradients (Maclean et al. 2003, Catford et al. 2012. The vertical vegetation structure hypothesis (MacArthur andMacArthur 1961, Recher 1969) can be tested in tandem with the intermediate disturbance hypothesis because the former may allow examining the mechanisms that underlie diversity patterns across disturbance gradients. The vertical vegetation structure hypothesis predicts that forest biodiversity will be greater in vertically complex stands than in structurally simple habitats Martin 2015a, Caviedes and. Avian community composition will vary differentially in forests with recent fires compared to older fires based on the trajectories of recovery of the system (Vierling and Lentile 2008, Clavero et al. 2011, Latif et al. 2016. In recently burned areas, some species benefit from the availability of new resources, e.g., aerial foragers and some woodpeckers (Kotliar et al. 2002), whereas other species will avoid burned habitats, e.g., Black-eared Miner (Manorina melanotis; Clarke et al. 2005). For example, nest density of the Black-backed Woodpecker (Picoides arcticus) was higher in the year following a fire compared to previous years, probably on account of the temporal increase in bark and woodboring beetles, which typically colonize recently burned forests (Nappi and Drapeau 2009). Over time, however, avian community composition changes, as has been shown for western boreal forests in Canada, where snag-dependent bird species decreased and shrub-breeding species increased after 25 years post-fire (Hannon and Drapeau 2005). After a fire, an initial period of higher bird species diversity can occur, followed by a decline when dead trees fall (Bock et al. 1978). Avian responses to burned sites may depend on the extent to which the ecological succession, and associated structural changes in the vegetation, gradually provide or cease to provide their life history requirements (Pons and Clavero 2010).
A functional guild is "a group of species that exploit the same class of environmental resources in a similar way" (Simberloff and Dayan 1991:115), and thus are expected to respond similarly to a given disturbance. The habitat attributes that form the structural complexity of the vegetation are essential if certain functional guilds of birds are to persist in altered sites (Wunderle 1997). Therefore, the effects of fire on habitat attributes may be especially important, because many birds respond strongly to these features in their post-fire habitat (Ganey et al. 1996). For example, the presence of habitat legacies, such as standing dead trees, coarse woody debris, and understory, allows the foraging and reproduction of certain guilds, such as large-tree users and understory users, in a post-fire site (Morissette et al. 2002, Whelan et al. 2002. In subtropical semi-arid ecosystems of Argentina, forests converted from structurally complex stands to dense grasslands showed an avian community dominated by habitat generalists and a significant reduction of nectarivores, foliage users, and bark insectivores because of the lack of trees and shrubs (Albanesi et al. 2013). When some pre-disturbance vegetation structure survives a fire, certain guilds of birds can remain or should be able to approximate to their pre-fire densities as compared to other guilds (Franklin et al. 2000, Whelan et al. 2002. In the face of climate change, fires are increasing in frequency in many ecoregions globally (Lindenmayer et al. 2008, Cochrane and Barber 2009, Driscoll et al. 2010. It is crucial to assess the effects of fire regimes on biodiversity, particularly in globally threatened ecosystems such as temperate forests of South America (Clavero et al. 2011). Temperate forests of Chile and Argentina have a relatively low richness of birds (51 species), but 41% of these are endemic (Vuilleumier 1985. Here, Araucaria (Araucaria araucana) forests have been shaped by a mixed-severity and low frequency natural fire regime, including surface and crown fires after volcanic eruptions, with widespread high-severity fires being infrequent (Arroyo et al. 1995, González et al. 2005. Araucaria forest cover, estimated to be 500,000 ha in the 1550s, has been reduced by almost 50% on account of intensive logging and fires (Lara et al. 1999). Mostly post-1880, the frequency of human-set fires in Araucaria forests has increased, resulting in severe effects on vegetation and soil (González et al. 2011, Perfetti-Bolaño et al. 2013, Zuñiga et al. 2021. Increases in burning, mainly associated with forest clearing, grazing, and logging activities of new settlers, could have strongly affected Araucaria landscapes, altering the vegetation mosaic and the fire regime (González et al. 2009). Although it is suspected that fires can have important consequences for forestdwelling biodiversity, little is known about how avian communities respond to the increasing frequency of fires in these ecosystems.
In this study, we examine the responses of avian diversity (richness and density) of functional guilds and habitat attributes in sites with different fire regimes (number of times a site has been burned and time elapsed since the last fire) in Araucaria forests of southern Chile. We evaluate the relationship between different habitat attributes and avian density, predicting that (1) slightly disturbed (i.e., sites burned once more than a decade ago) and moderately disturbed (i.e., sites burned once recently) habitats, will show greater diversity of species compared to unburned and highly disturbed (i.e., sites burned twice, with a recent fire event; intermediate disturbance hypothesis) habitats, and (2) avian diversity will be greater in vertically complex types of vegetation and lower in structurally simple habitats (vertical vegetation structure Hypothesis) because the former offers greater foraging and breeding opportunities for birds inhabiting these forests.

Study area
The study was conducted in the Araucarias Biosphere Reserve (ABR) in the Andean zone of La Araucanía Region, southern Chile (38º-39ºS, 71ºW; UNESCO 2010). We conducted the study in areas adjacent to the China Muerta National Reserve ( The fire in the Malleco National Reserve was caused by lightning, whereas all the others were caused by human accidental ignition. The climate is temperate with a short dry season (January-March) and an average annual rainfall of 1945 mm (Luebert and Pliscoff 2006). Elevation in the ABR ranges from 200 to 3747 m above sea level (m a.s.l.) with forests up to ~1700 m a.s.l. The Araucaria forest type is found at over ~1000 m a.s.l. (Donoso 1993). These forests are dominated by Araucaria araucana and species of the Nothofagus genus (Fuentes-Ramirez et al. 2018, Urrutia-Estrada et al. 2018, with Chusquea spp. dominating the understory (González et al. 2009).
We selected four treatments (hereafter, sites) according to their fire frequency and levels of disturbance: (A) control (unburned; undisturbed); (B) burned once in 2002 (hereafter, "burned 2002"; slightly disturbed because it was burned once and it experienced a relatively long time for recovery; sampled 16 years post-fire); (C) burned once in 2015 (hereafter, "burned 2015"; moderately disturbed because it was burned once and it experienced a relatively short time for recovery; sampled two years post-fire); and (D) burned twice, in 2002 and 2015 (hereafter, "burned 2002/2015"; highly disturbed because it was burned twice and it experienced a relatively short time for recovery; sampled three years after the last fire; Fig. 1). These four sites were between 1014 and 1469 m a.s.l. and separated by at least 5 km from the next closest site. In the three burned sites, we examined the severity of fire according to assessments of damage conducted by the National Forestry Service (CONAF 2017). The three sites suffered high-severity fires, i.e., crown fires with > 90% of trees and understory vegetation charred, in the surveyed areas. Furthermore, there was a dense layer of ash in the soil (up to 30 cm in depth) and large holes in the ground caused by tree stumps that were completely carbonized through below-ground combustion with evidence of tree carbonization, root damage, and significant loss of shrub vegetation (CONAF 2017, Zúñiga et al. 2021.

Habitat sampling
At each of our four sites, we systematically established 40 vegetation plots separated by at least 150 m (radius of 11.2 m; 0.04 ha; Total n = 160 plots; Ibarra and Martin 2015b). In each plot, we measured the following habitat attributes: density of live and standing dead trees (number/ha), diameter at breast height (DBH, in cm) of each tree, cover of bamboo understory (%), cover of coarse woody debris (%), cover of intermediate canopy (%), cover of upper canopy (%), and volume of coarse woody debris (m³; Ibarra et al. 2014, Caviedes and.

Bird sampling
In each site, we established 40 point counts at the center of each vegetation plot (Total n = 160 surveys). During two breeding seasons (December-February; Ralph et al. 1996), we sampled the control site and the "burned 2015" site in 2016-2017 and the "burned 2002" and "burned 2002/2015" sites in 2017-2018. Each station was surveyed once in our four sites. Each point count lasted six minutes during which all the birds seen and/or heard within a radius of 50 m we recorded and grouped into two distance intervals (0-25 m and 26-50 m; Bibby et al. 2000, Ibarra andMartin 2015b). For bird species with a low detection rate, we used call playbacks at odd-numbered point counts in order to avoid duplicate counting of species in neighboring points, i.e., 20 point counts with playback per site. Each call had a duration of 30 s for each species, followed by 30 s of silence between species' calls (total of 15 min of survey time for stations with playbacks). The order of playbacks was Magellanic Tapaculo SU G(H) G M † Based on primary use of forest structure for nesting and/or foraging. SU = Shrub user, VPG = Vertical profile generalist, LTU = Largetree user, UU = Understory user (Díaz et al. 2005). ‡ I = insectivore, G = granivore, N = nectarivore, F = frugivore, H = herbivore (Jaksic and Feinsinger 1991). Secondary foraging guild in brackets. Only the primary foraging guild was considered in the analysis. § G = ground, A = air, F = foliage, T = timber (Jaksic and Feinsinger 1991). Secondary foraging substrate in brackets. Only the primary foraging substrate was used in the analysis. | M = migrant, i.e., all the population leaves the area in winter. P = partial migrant, i.e., part of the population leaves the area in winter. R = resident, i.e., all the population remains in the area throughout the year (Jaksic and Feinsinger 1991).
We conducted the counts during the first four hours after sunrise (between 6:00 and 10:00 AM). We recorded temperature (°C), humidity (%), and wind speed (m/s) at each point count, using a portable meteorological unit (Kestrel 4200, Kestrel-meters, Birmingham, MI). We used a portable bluetooth speaker for playback calls.

Data analysis Habitat attributes, and avian detectability and density
We used one-way ANOVA with Tukey's post-hoc tests to analyze differences of habitat attributes between sites with different fire regimes. To estimate detectability (p) and density (D), we used only bird species recorded more than 10 times (n = 20 species, Table 1) at all counting points. Estimates of bird density vary depending on the detectability of the species and these may be affected by the distance of the observer (25-50 m) and other specific sampling covariates, including temporal factors and weather conditions (Buckland et al. 2001, Royle et al. 2004, Ibarra et al. 2014). To estimate p and D for each species at each point count, we used mixed multinomial Poisson models (Royle et al. 2004). Maximum likelihood methods were used in the R-Unmarked program (Chandler 2014) in R (R Development Core Team 2019). Density per ha of standing dead trees with DBH ≥ 12.5 cm 2.9 Volume of coarse woody debris (Cwd_vol) † Calculation based on length and diameter of each segment with diameter ≥ 7.5 cm, crossing the transect of 11.2 m in length, and/or standing dead trees with height ≤ 1.3 m (in N-S direction) † Covariates selected for testing temporal and environmental associations in birds in burned sites and control site after reducing collinearity.
Before estimating p and D, we evaluated collinearity to reduce the number of covariates. Collinear covariates (r > 0.6) were eliminated, retaining only those predicted to be the most biologically influential for each species (Table 2; Dainese andPoldini 2012, Ibarra andMartin 2015a). For each species, we first estimated p using temporal and weather covariates identified using AIC (Anderson et al. 2015). The models ≤ 2 AIC value units were considered as the competitive models and thus we averaged the estimates of p from these competitive models (ΔAIC < 2; Burnham and Anderson 2002). For each species, a step-wise covariate selection procedure was used (without parameterizing D) and each model was then classified by AIC, selecting the bestranked models to continue modeling D (Ibarra and Martin 2015a). We estimated D according to habitat attributes and, to obtain the best models of D, we created a set of candidate models based on the model weights (wi) and the accuracy of the estimated coefficients, using AIC. Models within 2 AIC units of the top model were considered as the competitive set of best-supported models (Burnham and Anderson 2002). An average range of 17 to 24 models of D was evaluated for the 20 species.

Functional guild responses to burned sites
Avian species were classified into guilds according to habitat use, foraging mode, foraging substrate and migratory status ( Table 1). The resultant guilds were related to the use of resources relevant to the ecosystem function, i.e., quantity, type, and temporality of the resource used and the strategies of each species to obtain the resources ( Table 2). Some of these guilds have been associated with the vulnerability of the species to habitat alteration in temperate forests (Díaz et al. 2005, Cofre et al. 2007, Ibarra and Martin 2015b. The response of the guilds to the different burned conditions was compared using generalized linear mixed models (GLMMs). These models describe the relationship between a response variable and a set of explanatory covariates (fixed effects) compiled in units aggregated at different levels (random effects). We tested the fixed effect of the control, "burned 2002," "burned 2015," and "burned 2002/2015" sites on the observed richness and estimated density of avian guilds (response variables). The sampling season (2016)(2017)(2017)(2018) was used as a random effect. For these analyses, the lme4 (Bates et al. 2015), nlme (Pinheiro et al. 2017), and AICcmodavg (Mazerolle 2017) statistical packages were used in R (R Development Core Team 2019).

Habitat attributes
The control site had the highest density of live trees (90% higher than the "burned 2002/2015" site; Table 3) and the highest volume of coarse woody debris (92% higher than the "burned 2002/2015" site). The "burned 2002" site had the largest average tree size (DBH; 76% higher than the "burned 2002/2015" site) and the highest percentage of understory cover (71.6% higher than the "burned 2015" site). The "burned 2015" site had the highest density of standing dead trees (96.5% higher than the control site). Out of all the sites, the "burned 2002/2015" site had the lowest density of live trees, the lowest average DBH, and the smallest volume of coarse woody debris ( Table 3).

Richness of bird community
We recorded 35 bird species (Fig. 2.1), corresponding to nine orders and 19 families. Most species belonged to the Passeriformes order (71.4%; n = 25). The Furnariidae and Tyrannidae were the most represented families, with seven and five species, respectively. The "burned 2002" site had the greatest avian richness (n = 24), whereas the "burned 2015" site had the lowest avian richness (n = 19). We recorded eight species exclusively in the burned sites, whereas nine species were recorded in all sites (Appendix 1). We recorded only two species, the Chilean Pigeon (Patagioenas araucana) and the Patagonian Tyrant (Colorhamphus parvirostris), exclusively in the control site.

Species detectability and density
The detectability of 12 of the 20 species was associated with temporal or climate covariates (Table 4). Cloud cover was the covariate associated with more avian species (n = 5). The estimated density of 17 species was associated with some environmental covariates (Table 4). For 13 species, we found a positive association with DBH, whereas the only negative association was for the Rufous-collared Sparrow (Zonotrichia capensis). The Magellanic Woodpecker (Campephilus magellanicus) showed the highest influence of DBH on its estimated density (Table 4). On the other hand, avian density differed among burned sites. The "burned 2002" site showed the highest bird density, which was 5.2 times higher than in the "burned 2002/2015" site, which showed the lowest densities ( Fig. 2.2).

Habitat-use guilds
The density of shrub users (SU; Fig. 3.1) was highest in the "burned 2002/2015" site (11.2 times higher than in the control site), whereas the highest density of vertical profile generalists (VPG; Fig. 3.2) was observed in the "burned 2015" site (7.5 times higher than in the control site and the "burned 2002/2015" site).
In the case of large-tree users (LTU; Fig. 2.3), density was highest in the "burned 2015" site (10.75 times higher than in the "burned 2002/2015" site). The lowest density of understory users (UU; Fig. 3.4) was in the "burned 2002/2015" site (Appendix 2).

Foraging guilds
The highest density of insectivorous birds (I; Fig. 4.1) was observed in the control site (six times higher than in the "burned 2002/2015" site), whereas the density of granivores (G; Fig. 4.2) was highest in the "burned 2002/2015" site (11.67 times higher than in the control site). In the case of nectarivores (N; Fig. 4.3), density was highest in the "burned 2002" site (1.2 times higher than in the control site), whereas frugivore densities (F; Fig. 4.4) were highest in the "burned 2002" site (1.9 times higher than in the control site; Appendix 2).

Foraging substrate guilds
The highest density of birds that use the ground as their foraging substrate (G; Fig. 5.1) was observed in the "burned 2002/2015" site (10.5 times higher than in the control site), whereas aerial foragers (A; Fig. 5.2), had the highest densities in the "burned 2015" site (0.3 times higher than in the control site) and was lowest in the "burned 2002/2015" site (control site was 3.85 times higher). The density of foliage gleaners (F; Fig. 5.3) was highest in the "burned 2002" site (9.7 times higher than in the control site). In the case of birds that use timber or bark (T; Fig. 5.4), density was highest in the control site (4.1 times higher than in the "burned 2002/2015" site; Appendix 2).

Migratory status guilds
The density of migrant birds (M; Fig. 6.1) was highest in the "burned 2002/2015" site (12.1 times higher than in the control site), whereas, for partial migrant birds (P; Fig. 6.2), it was highest in the "burned 2015" site (5.4 times higher than in the control site). The highest density of resident birds (R; Fig. 6.3) was observed in the control site (4.3 times higher than in the "burned 2002/2015" site; Appendix 2).

Response of the overall avian community to fire regimes
This study reveals important differences in the bird community responses to fire regimes in globally threatened Araucaria forests in the Andes of southern Chile. The results show variation in the Fig. 4. Values predicted by generalized linear mixed models (GLMMs) of density (birds/ha) for each point count of the avian foraging guild for (1) insectivores, (2) granivores, (3) nectarivores, and (4) frugivores in the four sampling sites: control (green), "burned 2002" (yellow), "burned 2015" (blue), and "burned 2002/2015" (red). Different letters indicate significant differences (p < 0.05) among treatments. The box indicates groups of numerical data through their quartiles, midline is the median of the data for each site, and the whiskers are the parallel lines that show the variability outside the upper and lower quartiles. Individual points that are beyond the whiskers indicate outliers that are numerically distant from the rest of the data. response of overall avian richness and density of different avian guilds, depending on the number of fires and the time interval between each fire of the surveyed sites.
We found that time since the fire likely influenced differences in avian community composition of each site (bird abundance and richness increased with stand age and vegetation growth), with results similar to those of Moreira et al. (2003) and Clavero et al. (2011). These two studies found that avian composition in a burned forest depends heavily on time since the fire, which is related to the progress of vegetation succession with an associated gradual increase in structural complexity. In a pine (Pinus halepensis) forest in Israel, avian diversity increased from forests burned one and five years previously (Izhaki and Adar 1997). This study also found that several species avoided burned sites in the first two years after the fire and only recolonized three to five years post-fire. The response of these species was strongly linked to vegetation succession over time in the burned forests (Izhaki and Adar 1997).
The highest richness and density of birds were observed in the "burned 2002" site, probably because of the partial recovery of the vegetation since the fire disturbance 16 years ago. These results partially support the intermediate disturbance hypothesis according to which there will be a greater diversity of species in slightly disturbed habitats (Connell 1978, Hidasi-Neto et al. Fig. 5. Values predicted by generalized linear mixed models (GLMMs) of density (birds/ha) for each point count the foraging substrate guild for (1) ground, (2) air, (3) foliage, and (4) timber in the four sampling sites: control (green), "burned 2002" (yellow), "burned 2015" (blue), and "burned 2002/2015" (red). Different letters indicate significant differences (p < 0.05) among treatments. The box indicates groups of numerical data through their quartiles, midline is the median of the data for each site, and the whiskers are the parallel lines that show the variability outside the upper and lower quartiles. 2012). The unpredictable nature of most fires, e.g., timing, location, and severity, often maintains landscape heterogeneity by creating a variety of successional stages; avian communities will vary among successional stages and forest types (Morissette et al. 2002). In the same Araucaria-Nothofagus forests sampled in this study, some native plants characteristic of the understory, such as Chusquea culeou, Gaultheria poeppigii, and Alstroemeria aurea, were able to re-sprout abundantly in burned areas (Urrutia-Estrada et al. 2018). In addition, A. araucana trees can reproduce from seeds or by re-sprouting after a fire (González et al. 2010, Fuentes-Ramírez et al. 2019, whereas N. pumilio reproduces only from obligate seed germination (Fuentes-Ramirez et al. 2018). This recovery, together with the mixture of early succession species and habitat legacies in the "burned 2002" site (slightly disturbed), could explain the greater richness of avian species we recorded in areas burned 16 years ago (Shiel and Burslem 2003).
The vertical vegetation structure hypothesis (MacArthur andMacArthur 1961, Recher 1969) was tested in tandem with the intermediate disturbance hypothesis in order to examine whether sites with greater structural complexity, e.g., relatively higher density of standing dead trees and a dense understory , host higher avian diversity than structurally simpler habitats. Our results are in accordance with previous studies conducted in other forest ecosystems, e.g., Brazilian Amazon and Australian forest, where the lowest avian richness was found in forests with greater frequency of fires (Peres et al. Fig. 6. Values estimated by generalized linear mixed models (GLMMs) of density (birds/ha) for each point count of migratory status species for (1) migrant birds, (2) partial migrant birds, and (3) resident birds in the four sampling sites: control (green), "burned 2002" (yellow), "burned 2015" (blue), and "burned 2002/2015" (red). Different letters indicate significant differences (p < 0.05) among treatments. The box indicates groups of numerical data through their quartiles, midline is the median of the data for each site, and the whiskers are the parallel lines that show the variability outside the upper and lower quartiles.
2003, Peres 2004, Lindenmayer et al. 2008). The latter can be explained by a reduction of the structural complexity of forest systems after fires (White et al. 2016).

Response of functional guilds to fire regimes
Avian guilds responded differently to the fire regimes of the different sites in our study. The density of all shrub user species was higher in all burned sites. This is consistent with the findings of Fontaine et al. (2009) andStephens et al. (2015), who observed that shrub users benefited from reduced tree cover and an increase in shrub volume up to nine years after the fire in a mixed broadleaved and coniferous forest in southwestern Oregon.
A higher density of large-tree users and vertical profile generalist birds was observed in the "burned 2002" (slightly disturbed) and "burned 2015" (moderately disturbed) sites, probably on account of the change in vegetation structure, e.g., increase of snags, which creates new foraging and nesting opportunities for the species in these guilds (Morissette et al. 2002). For example, large-tree users usually depend on decayed trees for nesting . The low density of large-tree users and vertical profile generalists in the twice burned site (highly disturbed) is likely associated with the low density of live and standing dead trees that are essential for several species from these two guilds (Perfetti-Bolaño et al. 2013). For birds that use old-growth trees as their foraging substrate, the highest density was observed in the control site (undisturbed). The rapid decline of these birds, such as the Magellanic Woodpecker and Chilean Flicker, in disturbed sites may be explained chiefly by the fact that these species require larger areas of forest and the presence of large decaying trees for feeding Martin 2015a, Vergara et al. 2016). Smucker et al. (2005) suggested that the response of a species to a fire is observed, in many cases, years after the event. For example, an excavator like the Northern Flicker (Colaptes auratus) showed an increase in density starting the third year after the fire (Smucker et al. 2005) but Wiebe (2014) found that the number of individuals increased in the year following the fire.
The low density of understory users in recently burned sites may reflect their association with a dense understory, promoting a negative response when the lower vegetation stratum is damaged by a fire (Reid et al. 2004). In addition, birds from the Rhinocryptidae family can use fallen trunks for refuge and reproduction (Venegas et al. 2009. The latter may explain the lowest density of these species in the twiceburned site in line with its lack of understory cover and refuge. However, we found that, as the understory recovers as reported for the "burned 2002" site (slightly disturbed), understory users can begin to recolonize an area in three to five years post-fire (Venegas et al. 2009). Many species of insectivorous birds (bark searching, arboreal, and terrestrial) avoid open areas and treefall gaps in burned forest-dominated landscapes (Barlow and Peres 2004). Kreisel and Stein (1999) observed that, after a fire in a coniferous forest in Washington, there was a decrease in the number of beetles under the bark and a decline in the abundance of four woodpecker species. In the Amazon, Barlow et al. (2002) found that both insectivores and bark-user birds were significantly less abundant in burned forests compared to unburned forests.
The greater density of granivorous and ground-foraging birds in burned areas could be attributed to greater access to seeds. Fire generally makes for an increase in the availability of some foraging resources, e.g., seeds of the herbaceous plants that colonize these sites (Milesi et al. 2002, Sosa et al. 2010). According to a study of a burned forest in central Chile, the rapid recovery of gramineousherbaceous and shrub cover resulted in a greater abundance of granivores (Perfetti-Bolaño et al. 2013). One year after a fire, the abundance of granivorous species was more than double that found the following year, suggesting that some species respond to short-term increases in seed availability after a fire (Smucker et al. 2005). For foliage feeders, our results differ from those reported by Dieni and Anderson (1999) and Grigera and Pavic (2007) who attributed the decrease in their density in post-fire sites to the reduction in horizontal and vertical foliage cover. A possible explanation for the increase in density we found in all the burned sites could lie on the effect of the White-crested Elaenia (Elaenia albiceps) on our results. White-crested Elaenia, a Neotropical austral migrant, is the most abundant bird species during spring and summer in the area and uses the foliage substrate for prey capture across the vertical profile (Chust et al. 2012). Large numbers of White-crested Elaenia influenced the density of foliage feeders in our study.
It has been suggested that migratory birds may be more resilient than resident birds to disturbances such as fires (Lynch and Whigham 1995). In South American temperate forests, almost half of the bird species are either partial migrants, i.e., part of the population migrates latitudinally, (31%) or total migrants, i.e., all the population migrates latitudinally (17%; Cofre et al. 2007. Petit et al. (1993) found that disturbed tropical habitats tend to contain more migratory forest specialist species than undisturbed tropical forests. This agrees with our observation of a higher density of migrant birds in the most disturbed sites ("burned 2002/2015"; highly disturbed), suggesting that some burned forests may still act as habitat for migrant bird species (Smucker et al. 2005).

Implications for avian conservation
Fires are likely to occur on an even larger spatial scale and with increasing frequency and severity as a result of climate change, significantly affecting previously unburned forests Schulze 1999, Peres et al. 2003). Thus, the dynamics of fire disturbance have now become a decisive factor in conservation and management of forest biodiversity. Although Chile currently lacks an adequate policy for the management of fires, national and local governments are exploring ways to prevent or at least minimize the threats posed by fires. Furthermore, Chile's policy for the protection of threatened species, such as A. araucana, demonstrates the country's will to ensure the protection of local biota. The possibility that fires allow the persistence and further recolonization of certain bird species and guilds should also be taken into account when designing post-fire management plans. This is especially important for current policies that encourage the implementation of "salvage logging" practices post-fire. Postfire management practices that affect the forest succession, e.g., planting exotic-trees and eliminating shrubs, or remove habitat legacies, e.g., harvesting standing dead trees for lumber or firewood, may be counterproductive for forest biodiversity conservation. No habitat-specialist bird or member of specialized avian guild that inhabits Chilean temperate forests has received either national or international conservation status or attention, except for C. magellanicus (Cofre et al. 2007). Furthermore, many forested areas in Chile have not been sufficiently surveyed and thus current ecological knowledge of specialized species in temperate forests is limited. Further research on how habitatspecialist species and guilds respond to fires in their population trends and productivity is needed in order to establish reliable conservation status assignments and stewardship priorities for forest biodiversity under rapid local, national, and international social-ecological changes (Ibarra et al. 2020). Responses to this article can be read online at: https://www.ace-eco.org/issues/responses.php /1999