Habitat loss and fragmentation are among the most evident threats to biodiversity and the related edge effects may drastically change ecosystems structure and function [1] -[6] . The Brazilian Atlantic forest, one of the world’s species-rich ecosystems [7] , was reduced to 11.7% of its original cover [8] and currently is restricted to remnant fragments [9] [10] that are potentially highly exposed to several threats. In this case, edge effects are important drivers of change when considering the physical and biotic alterations related to such artificial boundaries. Changes in resource availability, species interactions and disturbance patterns are expected in such human-modified habitats. Many studies have shown the consequences of edges effects in tropical forests [11] -[16] , but they are almost unknown the precise consequences on Atlantic forest regeneration patterns and, only recently, it was found that the creation of linear canopy openings caused edge effects on understory communities [17] .

Plant communities may also vary in impact resistance and resilience, and then in the human-induced complex scenario, it is also expected that local edge effects are site-specific and related to disturbance history. On the other hand linear canopy openings, such as gas pipeline and power lines are relatively controlled field experiments and allow us inferences about edge effects that are probably not confounded by other disturbances [17] [18] . Here we analyzed edge effect patterns on Atlantic forest woody seedlings. In this study, seedling communities at forest edges, adjacent to the linear canopy openings, and thus gas pipeline and power line, were compared with forest interiors used as control sites.

The GP and PL edges always showed lower seedling densities when compared to the INTERIOR. In the years 2000 and 2001, recruitment was extremely low in edge sites, even excluding the expected pattern of high mortality of newly germinated seedlings. However, the mean basal area was similar between the sites, suggesting that the local seedlings are able to resprout or resist even facing the edge effects. Moreover, seedlings with more than 4 mm diameter apparently were more resistant to direct damage or edge effects and this fact was very important to the basal area patterns (Table 1).

Edge effects are important ecological filters for plant establishment, survival, and growth [11] [23] and here the edge sites, thus, gas pipeline (GP) and power line (PL), showed less favorable conditions for establishment and initial survival of seedlings. Some years after the fragmentation, in the phase called “stagnation”, seedling densities at the edges tend to be lower when compared to the forest interior [24] -[28] . These reductions may be caused by several factors such as the small recruitment [29] which can be originated from biotic [26] [30] or abiotic causes mainly derived from changes in disturbance regimes, species interactions and resource availability [2] [14] [31] .

Nevertheless, lower densities on edges may also be caused by limitations on the arrival of seeds. On tropical rainforests the dispersal in space, mediated by animals, may be negatively affected by the fragmentation [32] -[35] . In small forest fragments may also be a lower recruitment of seedlings of animal dispersed species [32] . In fact in some Atlantic Forest areas, an interruption in the dispersal processes by vertebrates may lead to a dominance of pioneer species with small seeds especially on edges [36] . The pattern of more abundance of disturbance-tolerant species in edges was previously observed for understorey plants [17] and trees [18] at the União Biological Reserve. Therefore, these linear canopy openings may contribute to change continually and permanently the patterns of occurrence of dispersal fauna, resulting in changes in the arrival of seeds in the adjacent forests.

The small seedlings are very susceptible to disturbances [37] [38] then it would be expected that these individuals had higher mortality [39] . On the other hand, observing the habitat, younger edges and contrasting limits (edge vs. matrix) in general have more intense effects on remnants [11] [40] [41] . Thus, gas pipeline (GP) showed the most unfavorable condition for seedlings with observed high mortality and lower recruitment.

Linear canopy openings may cause edge effects, promoting low seedling densities on edges by the reduction on germination and establishment. Therefore, these types of edge effects are distinct from those observed in typicaltropical forest fragments where the low densities are caused mainly by increase in mortality.

Data without normal distribution were transformed to log₁₀(x + 1). For unequal n’s the Tukey test used was Spojotvoll/Stoline. ^*Recruitment and ^*Deads are based on one year period (2000 to 2001 year).

We thank Dr. Andrea Costa and one anonymous referee for valuable suggestions which improved the manu- script. We are very grateful to Adilson da Silva and Adilson Pintor for fieldwork support. We also thank the Fundação Flora and PETROBRAS for financial support (research grant number 6000.0023998.06.02 to Programa Mata Atlântica/Pablo J.F.P. Rodrigues), the Instituto Chico Mendes de Conservação da Biodiversidade (ICMBio) and the União Biological reserve staff for logistical facilities, A. Costa for linguistic advice, and UENF/FENORTE for granting a Phd’s fellowship to P.J.F.P. Rodrigues.