Searching for bioindicators of forest fragmentation : passerine birds in the Atlantic forest of southeastern Brazil

Aiming to evaluate the potential of Passerine birds as bioindicators of forest fragmentation, we studied the avifauna in the mountain region of the state of Rio de Janeiro by mist-netting between 2001 and 2005. We sampled six sites, including four small fragments (from 4 to 64 ha) in an agricultural area (Teresópolis), one second-growth forest (440 ha Miguel Pereira) and a continuous forest (10,600 ha, Serra dos Órgãos National Park SONP). Indicator Species analysis and a Monte Carlo test were run to detect associations between species and sites, considering at least 30% of perfect indication and a significant value for the statistical test, only considering species with at least 10 captures. A total of 30 Passerine birds were sampled, and due to their association to the largest area (SONP), we considered Sclerurus scansor, Mionectes rufiventris, Chiroxiphia caudata and Habia rubica as the best indicators for this area. Five species were more captured not by chance in the 440 ha second-growth: Conopophaga melanops, Myiobius barbatus, Myrmeciza loricata, Philydor atricapillus and Schiffornis virescens and no species were related to any small fragment. This analysis has identified specialized Passerine species in many aspects (foraging substract, needs for nesting, rarity status), and it can be a valuable tool for detecting possible bioindicators.


Introduction
Ecological indicators have been used to detect changes in nature for the last 40 years, and currently, they are mainly used to assess the condition of the environment (Niemi and McDonald, 2004), and one of the aspects that have been evaluated by ecological indicators is the landscape fragmentation (O'Neill et al., 1988).
The vegetation in the four smaller fragments, from now named F1 to F4 (from smaller to larger), is secondgrowth forest.They are surrounded by pastureland for raising cattle (which usually is seen inside the fragments), and agricultural activities (in which pesticides are used).Excepted for F1 (where only one net line was operated), we performed a two-net lines protocol with from 6 to 10 mist nets in each fragment.In every fragment, one net line was close to the border and the other was located inside the forest.Every fragment and net line was sampled at least every two months, with the same sample effort in all studied sites.
Igapira Farm is a 910 hectares private farm in the municipality of Miguel Pereira.It is estimated that 440 hectares are covered by second-growth montane pluvial forest, and the remnant are buildings and large area of pasturelands for cattle and buffalo.The dominant native covering is montane semideciduous Atlantic Forest (Rizzini, 1997), with trees between 20 and 30 m, and emergent reaching up to 40 m.For sampling birds, eight to ten mist nets were operated every month in two sites (800 and 1,100 m high) The vegetation in SONP is second-growth Atlantic pluvial forest, with palm trees, epiphytes and tall trees.It has 10 peaks higher than 2000 m and six other peaks over 1500 m high.The highest peak is Pedra do Sino at 2263 m, and the lowest is located in the relatively flat municipality of Magé, at 145 m.The average temperature at 1000 m is 17 °C (Wikipedia, 2006).We sampled two sites, located at 1000 m high and about 2 km far from each other.From 6 to 10 mist nets were used in each site, which was sampled every two months.
Due to different mist netting efforts (net-hours) in the six sampled areas (850 in each fragment, 2680 in Igapira Farm and 1800 in SONP), before using our data we divided captures by net-hours x 100; thereby we were able to compare our results of relative proportion of captures in these sites.Mist nets are probably free of biases associated with the observer (Bierregaard and Stouffer, 1997); however introduction biases themselves in that birds of different sizes and with differing behavior patterns, whose capture probabilities are distinct (Remsen and Good, 1996).Nevertheless, their use has provided satisfactory results in many studies (e.g., Bierregaard and Stouffer, 1997;Blake and Loiselle, 2001;Wang and Finch, 2002;Piratelli and Blake, 2006), and this methodology can be an useful tool for comparing bird communities of several sites when used in a similar sampling effort.
The Indicator Species analysis (Dufrene and Legendre, 1997) and a Monte Carlo test were run using the package PCOrd 4.01 (McCune and Mefford, 1997) (Mittermeier et al., 1999), mostly in small fragments.The Atlantic Forest harbors 688 breeding species of birds (Goerck, 1997), only one bird species is truly registered as extinct in this ecosystem, but many species are now endangered (Collar et al., 1992).In the state of Rio de Janeiro, the most important remnants of this ecosystem are in the mountains, and mainly restricted to fragments.This state hosts 653 taxa of birds, including 82 threatened, 38 probably threatened and 25 without enough data for evaluation of their status (Alves et al., 2000).
One of the primary goals of research on bioindicators is to identify species or other taxonomic units that would indicate disturbances in the environment, reflecting the responses of other species (Rainio and Niemelä, 2003).When searching for bioindicators, one has to find groups of species which respond quickly to environmental changes, being these answers clearly distinct from those generated by natural fluctuations (Furness and Greenwood, 1993).Birds have been used as bioindicators for having particular characteristics, as good taxonomic and systematic knowledge (Furness and Greenwood, 1993;Bierregaard and Stouffer, 1997), several species occupying higher levels on food chain (Bierregaard, 1990) and sensibility to habitat losses and fragmentation (Terborgh, 1977;Turner, 1996).
In fragmented landscapes, a regular process is the substitution of rare-specialized species for abundant-generalist ones, more adapted to open and clearcut areas (Willis, 1979;Restrepo and Gómez, 1998;Willis and Oniki, 2002a); thus we can associate the presence/absence of some species to environmental quality.In general, Suboscines Passerine birds are more closelyrelated to forested landscapes, and Oscines usually are more linked to open areas (Willis, 1979;Sick, 1997).If those assumptions are natural processes, some species usually would be more easily found in continuous forests; meanwhile other groups will appear mainly in open and fragmented areas.Therefore, this non-random distribution would be a clue for evaluating environmental quality.
Our goal in this study is to verify non-random distribution of some Passerine bird species in areas with different levels of disturbance at the Serra do Mar sub region of Atlantic Forest.Thus we intend to detect which species are more closed-related to each site (indicator species), investigating their potential as bioindicators of forest fragmentation.

Material and Methods
In order to estimate the potential of some Passerine birds as bioindicators of forest fragmentation in the Atlantic forest, we have conducted this research in the mountains of Rio de Janeiro state, southeastern Brazil, from October 2001 to July 2005.We used mist nets for sampling Passerine birds in six sites, including four small fragments (4, 9, 23 and 64 ha) in agricultural areas in Teresópolis municipality (22° 17' 22" S and 42° 52' 3 " W and 22° 16' 35" S and et al., 1999), nesting in holes in banks inside the forest (Develey and Endrigo, 2004).It has been considered a rare, specialized (Goerck, 1997) and highly sensitive species (Ribon et al., 2003), commonly as a member of fixed flocks (Develey and Peres, 2000).There are records of S. scansor in a 230 ha patch of semideciduous forest in cane fields of central São Paulo (Willis and Oniki, 2002b).These last authors did not find this species in further samplings, suggesting its disappearance due to loss of the few individuals present earlier.Even though it has already been observed in very small fragments by D' Ângelo-Neto et al. (1998) in Minas Gerais (5 to 8 ha) and in Rio Grande do Sul, foraging on the ground (I.Accordi, p. comm.), its abundance estimate was affected by logging in São Paulo, occurring mainly in a primary forest (Aleixo, 1999).
Mionectes rufiventris is a common species in lower growth of humid forest, secondary woodland and borders of southeastern Brazil (Ridgely and Tudor, 1994a) occurring in mixed flocks (Develey and Peres, 2000) both in lowland (Machado and Fonseca, 2000) and montane evergreen forests (Brooks et al., 1999).Aguilar et al. (2000) found 19 nests of this species in two of the three fragments (50, 200 and 300 ha) that they studied in Minas Gerais; however, they did not specify in which fragment the nests were found.All nests were constructed over water, fixed to tree roots under stream beds; the authors considered this species as nest-site specialist.Pizo and Aleixo (1998) has described its leck behavior in a 2000 ha protected area in São Paulo.There are records for this species in antropized areas in Minas Gerais (Manhães and Ribeiro, 2005) as well as in larger patches in Rio Grande do Sul (I.Accordi, p. comm.).
Chiroxiphia caudata is a common species in lower and middle growth of humid forest, secondary woodland, and borders (Brooks et al., 1999;Ridgely and Tudor, 1994a).It has been reported in a large fragment (21,840 ha) (Donatelli et al., 2004), and in smaller ones in São Paulo (Willis and Oniki, 2002b;Poza and Pires, 2003) and Minas Gerais (D'Ângelo-Neto et al., 1998;Ribon et al., 2004).In this last state, there are reports up to 500 m high in the Rio Doce Valley (Machado and Fonseca, 2000), and from 800 to 1,000 m high in Serra do Cipó ( Mello-Júnior et al., 2001).It was not affected by logging in São Paulo, occurring both in a primary forest and in a logged area (Aleixo, 1999).
Habia rubica lives in lower growth of humid forest and forest borders, usually in mixed flocks (Ridgely and Tudor, 1994b).It is a low specialized species (Develey and Peres, 2000) having medium sensitivity (Ribon et al., 2003).There are records of this species in a 21,840 ha forest (Donatelli et al., 2004), and in smaller fragments (Willis and Oniki, 2002b) in São Paulo.It is currently seen as a nuclear species of mixed flocks in Rio Grande do Sul (I.Accordi, p. com.).
Those four species, regardless of their records in small fragments, have some ecological characteristics (e.g.rarity status, needs for nesting, foraging in mixed to verify non-random species distribution in the sampled sites.We only considered species with 10 or more captures, a minimum of 30% of perfect indication and a significant value for the statistical test.We then assumed that those species more closely related to a given sampled site would be more specialized and probably more sensitive to the degradation of their habitats.Next, we searched in the literature for their biological characteristics, based on descriptions of their occurrence.Finally, we matched our results with previous descriptions in order to find the best bioindicators, and to test the use of Indicator Species analysis for detecting bioindicators.
In this analysis some species, even with an expressive higher relative abundance in a given sampled site were not considered as indicator, due to their low frequencies of captures.This is the case of Euphonia pectoralis (Latham, 1801) and Saltator similis d'Orbigny & Lafresnaye, 1837 for SONP (Table 3).
Sclerurus scansor usually is found on or near ground in humid lowland and montane forests (Brooks x 100 and trophic guilds (according to Willis, 1979 andSick, 1997).Table 1.Continued... flocks, sometimes as nuclear species) which would limit their occurrence and/or their nesting activities, leading them to prefer less altered areas.Thus, their indication for the largest area (SONP) in our analysis has matched their ecological specialization.

Families
Conopophaga melanops is a usual member of mixed flocks (Develey and Peres, 2000) in lowland forests (Brooks et al., 1999).Machado and Fonseca (2000) captured this species up to 500 m high in the Rio Doce Valley, Minas Gerais.It also was not affected by logging in São Paulo, occurring in the same areas described before for C. caudata (Aleixo, 1999).
Myiobius barbatus is an uncommon to fairly common species in lower growth of humid forest and ma-ture secondary woodlands (Ridgely and Tudor, 1994a).Machado and Fonseca (2002) sampled this species in the highest area of Rio Doce Valle.Its estimate of abundance was affected by logging in São Paulo, occurring mainly in a primary forest (Aleixo, 1999).
Myrmeciza loricata is a lowland and montane forest species (Brooks et al., 1999), living, for example, in the lower area of the Rio Doce Valle.(Machado and Fonseca, 2000) and having medium sensitivity (Ribon et al., 2003).
Philydor atricapillus is an inconspicuous bird in lower growth of humid forest and mature secondary woodland (Ridgely and Tudor, 1994a).It is a mediumsensibility species (Ribon et al., 2003) typical of mixedflocks (Develey and Peres, 2000).These last authors    have reported young individuals in an 80,000 ha protected area in São Paulo; in this state it also was not affected by logging (Aleixo, 1999).It is critically endangered in Rio Grande do Sul, occurring only in a few fragments of lowland forests in Northeastern, following Habia rubica (I.Accordy, p. com.).Schiffornis virescens is common in montane secondary forest (Brooks et al., 1999) and in mature secondary woodlands (Ridgely and Tudor, 1994a).It was recorded in São Paulo (Willis and Oniki, 2002b;Donatelli et al., 2004) in areas ranging in size from 230 to over 21,000 ha).Aleixo (1999) detected the influence of logging in this species in that state.
The last five species also have some characteristics (e.g.foraging in mixed flocks, medium sensibility, reports for very large areas, been affected by logging) that would restrict their occurrence to medium to large area and/or mature secondary woodlots.Again Indicator Species analysis has detected some sensitive species.
Because indicator species analysis combines both species relative abundance and frequency of occurrence (Dufrene and Legendre, 1997), it can indicate which species are more closely associated to specifics habitats not by chance.Ecological indicators should trade off two potential endpoints; they should be sensitive enough to react in a detectable way when a system is affected by anthropogenic stress and they should also remain reasonably predictable in unperturbed ecosystems (Niemy and McDonald, 2004).
In accordance with Stotz et al. (1996), who have suggested that for guiding principles for bird conservation in the Neotropics, one might focus on assemblages of endemics and habitat specialists rather than on sheer number of species, we do suggest using Indicator Species analysis as a clue for detecting groups of Passerine bird species that would be bioindicators.This could be a valuable tool for a quick and primarily evaluation of the environmental quality (mainly focusing on habitat fragmentation) and for bird conservation as well.

Table 1 .
All Passerine birds (by Families) sampled in six sites at Atlantic Forest in the state of Rio de Janeiro, Brazil, their capture rates -(captures/net-hour)

Table 2 .
Indicator values (% of perfect indication, based on combining values for relative abundance and relative frequency) and Monte Carlo test of significance of observed maximum indicator value.Only bird species with ten or more captures were considered in this analysis.

Table 3 .
Relative abundance, frequency of captures (in parenthesis) and percentile of perfect indication (when statistically significant) of birds with ten or more captures in six sites at Atlantic Forest in the state of Rio de Janeiro, Brazil.