MULTI-SPATIOTEMPORAL SIMULATION OF EDGE EFFECT ON FOREST PATCHES IN THE BARRA SECA RIVER BASIN, ES

The Atlantic Forest is intensely fragmented and this fragmentation process has caused an expressive increase of forest remnants and, consequently, increased edge effect with different physical-biological intensities in the transition areas between the patch and the matrix. This study used landscape metrics to understand and analyze how different edge effect distances affect the structure of the forest landscape in the Barra Seca River basin (ES), in 1985, 1996, 2006 and 2016. Remote sensing images were processed and using the Bhattacharya algorithm with supervised classification, the forest patches of the study area were classified and isolated. Landscape ecology metrics were computed with Patch Analyst and V-Late 2 Beta extensions. The forest patches were divided into four size classes as follows smaller than 5 ha (C1); between 5 and 10 ha (C2); between 10 and 100 ha (C3); and over 100 ha (C4). The edge effect simulation using landscape metrics was performed using the edge effect distances of 20, 40, 60, 80, 100, 140, and 200 m. Forest fragmentation increased between 1985 and 2016 while the number of patches greater than 100 ha decreased. Currently, the basin landscape consists mainly of small patches, which have larger relative areas affected by edge effect while many patches smaller than 10 ha are completely dominated by edge effect for distances greater than 60 meters. The edge effect simulation for different distances allowed verifying the intensification of the edge effect on the forest patches of the Barra Seca River basin.


INTRODUCTION
The Atlantic Forest is considered a priority area for conservation due to its high biodiversity and threat level (MYERS et al., 2000). The main threat against this biome is caused by disturbances of natural ecosystems that result mainly from anthropic processes types and the resulting forest fragmentation, one of the most significant threats to biodiversity (HAGEN et al., 2012).
Only 11.26% of the original Atlantic Forest cover remains while the degradation of this biome keeps growing at an accelerated rate (RIBEIRO et al., 2009). In Espírito Santo, the Atlantic Forest remnants account for only 9% of the original area, with a history of fragmentation and degradation that became more accentuated in the 1960s because this biome was the largest source of wood for the construction of Brasília (THOMAZ, 2010). Furthermore, even small size forest patches may have a significant number of species of fauna and flora, constituting priority areas for conservation (HAGEN et al., 2012).
The decreasing size of forest remnants caused by the fragmentation process changes the biotic community and the patch physical parameters. These changes act more intensely on the patch edges, affecting differently the biodiversity and the area under edge effect influence while depending on several factors, such as biome permeability, environment (matrix) characteristics, species sensitivity to changes, distance between forest patches and their dimensions (PRIMACK; RODRIGUES, 2001).
The edge effect, therefore, changes the structure and dynamics of ecosystems and is one of the main consequences of forest fragmentation. Thus, a quantitative analysis of the landscape structure using landscape metrics helps to understand such changes (ZARAGOZÍ et al., 2012).
The intensity of the processes related to the physical and biogeographic changes that act on the patch varies according to its dimensions (PRIMACK; RODRIGUES, 2001). To this end, it becomes important to study forest patches by size classes to understand better the evolutionary dynamics that shaped the current landscape.
The Barra Seca River basin has one of the few continuous remnants of the Atlantic Forest in Espírito Santo, and it is one of the 14 centers with high plant diversity in Brazil (PEIXOTO; SILVA, 1997). Thus, the objective of this study is to investigate how the edge effect distance affects the forest patches of the Barra Seca River basin (ES), using remote sensing techniques and landscape metrics.

Study site
The Southeast Atlantic Region houses the hydrographic basin of the Barra Seca River, which flows into the sea between Linhares and São Mateus, in Espírito Santo. Although this basin is considered a subunit of the Doce River hydrographic system, the waters have no direct contact ( Figure 1).
The Barra Seca River basin covers an area of 2216.56 km² and all the coastal sub-basins that occur in the area of quaternary accumulation, immediately north of the Doce River mouth.
In the 1985 and 2016 period, the increase of areas smaller than 100 ha was associated with an increasing number of forest patches in these classes and decreasing number of patches larger than 100 ha in C4. Class C1 increased 39.3%, from 1294 patches in 1985 to 1803 in 2016 while the C4 class (> 100 ha) decreased 14.8%, from 88 fragments in 1985 to only 75 in 2016 (Table 2). The high number of patches in the very small and small size classes, C1 and C2, highlights the predominance of very small areas for the maintenance of the biota of the basin. Despite the small number of patches in class C4 compared to other size classes over the period, the total core area of class C4 is significant compared to the total core areas of the other classes, emphasizing the importance of class C4 to the maintenance of biodiversity in the Barra Seca River basin.
The total core area of forest patches for the years 1985, 1996, 2006 and 2016, are shown in Table 3 Figure 2 shows the behavior of the total core area index (TCAI) and mean core area (MCA) obtained for the different edge effect distances. The total core area index (TCAI) decreased markedly in all smaller classes (<100 ha; C1, C2, and C3) for edge effect distances of 40 m and higher, during all evaluated years. Furthermore, over the 80 m edge effect distance, the relative core area of the classes smaller than 10 ha (C1 and C2) is close to zero, demonstrating that for small forest fragments at this given distance, the area without influence of edge effect is practically nonexistent.
For patches larger than 100 ha (C4), the core areas show a slight gradual decrease with increasing edge distance. Only in this size class, the farthest edge effect distance of 200 m had nearly no effect on forest patches, demonstrating that only the C4 class has an area capable of supporting species sensitive to the edge effect for the 200 m distance.
The mean core area (MCA) shows an upward curvature and much higher values in class C4 (> 100 ha) compared to the other classes. The curve shape shows that as the edge effect distance increased, the MCA also increased for the distances of 100 m and over in 1985 and 1996, 60 m and over in 2006, and 80 m and over in 2016 (Figure 2). The MCA increase that is observed from a certain edge distance onward is related to a decrease in the number of core patches, which follows the same trend as the MCA index, therefore, the total core area (TCA) is divided by a lower number of core patches, thus increasing the mean value.

DISCUSSION
The forest patches of the Barra Seca River basin are highly fragmented, constituted mainly by small size patches, and undergoing a continuous process of fragmentation in the studied time interval. Forest fragmentation causes the forest area to decrease while simultaneously increasing edge effect and dividing large remnants into small and non-continuous fragments and the influence of distance of isolation and time since isolation are associated, resulting in variation in response of the different species and populations (ROGAN; LACHER JR. 2018).
The intense fragmentation of the Atlantic Forest remnants is verified along the whole extension while most of the forest patches (83.4%) are smaller than 50 ha and distributed in a landscape of low connectivity between forest patches, with average isolation distance of 1400 m that increases to 3500 m for patches larger than 50 ha (RIBEIRO et al., 2009).
Large forest patches are very important for the maintenance of both the biotic community and ecological processes. On the other hand, small patches are less rich and may have insufficient areas to house viable populations of many species present in the study area that are at risk of extinction such as Panthera onca, Puma concolor, Tapirus terrestris, Priodontes giganteus, and several others (SRBEK-ARAUJO et al., 2014). However, small patches are crucial because they act as a complement to habitat and refuge areas for the broad-lived biota while reducing the degree of isolation of the larger patches and consisting of important ecological stepping stones (ROGAN; LACHER JR. 2018).
The ecological quality of the patches can be determined more efficiently by the core area compared to only the total area (MCGARIGAL; MARKS, 1994). Therefore, a fragment with large total area, but with influence of the matrix over the entire area does not harbor species that are sensitive to the edge effect (TURNER; GARDNER, 1990). Thus, the extension of the edge zone (the region under the matrix influence) is crucial for the existence of interior habitats in forest fragments (MATLACK, 1994). Several studies demonstrate an intense fragmentation of the Atlantic Forest and predominance of small patches (CEMIN et al. 2009;ZANELLA et al., 2012;PIROVANI et al. 2014).
The similarity analysis of the behavior of the metrics indices allowed clustering forest fragments into two groups, one smaller (C1, C2, and C3) and another larger (C4) than 100 ha. The analyzed core area indices decreased markedly for fragments smaller than 100 ha from the edge effect distance of 40 m onward. In addition, the total core area (TCA), total core area index (TCAI) and mean core area (MCA) decreased with increasing edge distance yearly and over the studied time interval, except for TCA. TCA decreased with increasing edge distance yearly between 1985 and 2016, but increased over the studied time interval.
For forest fragments larger than 100 ha, the TCA and the TCAI decreased with increasing distance of edge effect and sharply from the edge distance of 100 m in the studied period. The MCA index behaved similarly to other classes, reducing sharply from the edge distance of 40 m; however, it increased between the edge effect distance of 100 and 200 m, in the 1985 and 2016 interval. This result can be explained by the decreasing number of core areas from the edge effect distance of 80 m when the number of core areas is higher in 1985 compared to 2016.
The tendency of core area to increase for small fragments (<100 ha) and decrease for large fragments, is related to the increasing forest fragmentation of the study area, which creates a large number of small fragments thus increasing the total area of these fragments.
Many studies have demonstrated that edge effect acts on reducing biodiversity, which is caused by physical-biological changes and the sensitivity of species to the changing characteristics of forest fragments (CHIARELLO, 1999;PARDINI et al., 2009;LIMA et al., 2015FREITAS et al., 2018. Several actions can be taken to mitigate edge effects, such as the management of the fragmented landscape with forest restoration alternatives, which aim at increasing the fragment area and protecting the edges as to maximize the biological flow (ZANELLA et al., 2012).
The edge should be protected with fast-growing species (CALEGARI et al., 2010), preferably native, which already interact with the local biota. In addition, agroforestry systems consisting of a consortium of tree species and herbaceous species can be used as well, working as buffer between protected areas and intensively managed area (PARDINI et al., 2009;ASASE;TETTEH, 2010). Another suggestion is the encouragement of silviculture, especially with species that require low biocide management, which favor the permeability of the fauna, thus increasing the connectivity between fragments (BAKER et al 2013), such as Eucalyptus and Hevea brasiliensis (seringueira),