FOREST ANALYSIS IN THE LANDSCAPE CONTEXT AT SÃO BARTOLOMEU RIVER BASIN, VIÇOSA-MG

The reduction of forest areas in recent years has infl uenced landscape fragmentation, compromising ecosystem functions. The natural forest, in addition to maintaining biodiversity, is capable of improving the hydrological conditions of the watershed. This study aimed to characterize the landscape structure of the remaining forest fragments in the São Bartolomeu river basin, at the Zona da Mata of Minas Gerais State, and to assess their physiographic and pedological representativeness. The study area is responsible for most of Viçosa’s water supply, while it is undergoing an urbanization process, reinforcing the need for ordering the use, besides the evaluation of priority areas for conservation. The studies related to the landscape structure were based on the metrics of forest fragments. The representativeness was analyzed regarding the aspects of topographic compartmentation, slope faces, and soils, using the Chi-square test. The natural forest presents fragments varying from medium to large size, with satisfactory proximity between them. However, the forest patches have irregular shapes, favoring the edge eff ect with a tendency towards irregularity as the fragment size increases. The results of the physiographic and pedological representativeness indicate that the land uses are dependent on the location in the landscape. According to the statistical test, the forest occurs less frequently than expected on the river terraces (48.3%) and convex slopes (2.3%), as well as on the Red-Yellow Acrisols (53.1%), and on the north (28.5%) and west (20.9%) faces. For urban planning and biodiversity conservation purposes, it is recommended to balance the forest area in the diff erent compartments of the landscape, giving priority to the connection of forest fragments.


INTRODUCTION
The forest fragmentation process is associated with land use changes, dictated mainly by urbanization and agricultural occupation. In most landscapes, the sustainable limit of forest fragmentation has already been exceeded and the connectivity reduced. In these cases, the richness and abundance of local species become dependent on the size and environmental quality of the remaining fragments (Pardini et al., 2010).
The reduction of sites with landscape-specifi c pedological and physiographic characteristics is one of the consequences of forest fragmentation (Sarcinelli et al., 2012). Then, the organisms are generally in restricted and specifi c environments, resulting in landscapes with little habitat diversity (Fahrig, 2003). Therefore, topographic and edaphic variations play an important role in the spatial organization of tree vegetation on a local scale (Martins et al., 2003).
The isolation of the remaining forests fragments hinders the dispersion of the species, reducing the gene fl ow, genetic variability, and adaptability of these species (Guariz and Guariz, 2020). To mitigate these eff ects, it is essential to develop research and actions that result in practices for the development of a sustainable environment.
Landscape characterization applies several methods and techniques of metrics to analyze the landscape and can give important insights for the adequate management of natural resources (França et al., 2019). For instance, the representativeness of the environmental diversity (Sarcinelli et al., 2012) is used better understand the dynamics of the landscape structures. This approach is based on ecosystem ecology and spatial modeling and had a major advance with the rapid development of Remote Sensing and Geographic Information Systems.
Landscape-based conservation studies are essential for various purposes, such as delimitation of ecological corridors and landscape connectivity (Guariz and Guariz, 2020;Salomão et al., 2018;Schackelford et al., 2018); analysis of spatial planning and protection of environmental services (Inkoom et al., 2018); recovery of degraded areas (Paula et al, 2015;Toledo et al., 2018); wildlife management (Kosicki, 2018;Morelli et al., 2018); and planning of urban areas (Ribeiro et al., 2020). Therefore, these studies give the basis for proper management of resources within the context of the landscape.
Currently, the São Bartolomeu river is responsible for much of the water supply of about 70% of the municipality's population in the rainy season and 30% in the dry season of Viçosa, located in the Zona da Mata of Minas Gerais State (Bastos et al., 2009). The trend of urban expansion in the basin can accelerate the process of forest fragmentation if environmental criteria in urban planning are not well defi ned. Then, the adoption of measures to preserve biodiversity and the proper functioning of the region's water system are essential for human well being in this basin. This study aimed to characterize the structure of the remaining forest fragments and to evaluate their physiographic and pedological representativeness, and support the defi nition of priority areas for forest conservation.

Study area
The São Bartolomeu river basin is located in the municipality of Viçosa, Zona da Mata de Minas Gerais, between the parallels 20º44' and 20º51' South Latitude and meridians 42º55' and 42º50' Longitude West of Greenwich ( Figure 1). This basin is located in the Atlantic Forest biome and covers about 2,800 hectares, with an altitude ranging from 655 to 894 m. The Montana Semidecidual Seasonal Forest is the main forest natural typology (Veloso et al., 1991), and the region is characterized by intense agricultural activity developed by small farmers and, more recently, urban expansion, with the implementation of condominiums and allotments. The climate is Cwa type, with rainy summers and cold and dry winters (Martins et al., 2018). The geology in this study region is characterized by the predominance of graniticgnaissic substrate (Ab'Saber, 1970), and geoforms with an asymmetric concave-convex-top sequence and steep part of the slopes (Rezende, 1971).
For this study, we used the land use land cover map developed by Moreira (2009) from the visual classifi cation of Ikonos II image georeferenced and orthorectifi ed with PEC Class A for the scale of 1:10,000 (Santos, 2008). The land use land cover classifi cation resulted in nine categories: agriculture, urban area, rural buildings, coff ee, native forest, planted forest, hydrography, pasture, and pathways. The scale used in this work allows mapping fragments smaller than 1 ha accurately.

Description of landscape structure
Using the land use land cover maps, we analyzed the landscape structure based on several metrics of the forest fragments: total class area (CA), average fragment size (MPS) -including standard deviation (PSSD) and coeffi cient of variation of sizes (PSCoV) -, number of fragments (NumP), shape index (MSI), Area-Weighted Mean Shape Index (AWMSI), total edge (TE), edge density (ED), proximity indexes (MPI, with distance of 500 m, and MNN) and central area index (TCAI). These metrics were derived using the Patch Analyst extension of Arcgis 10.1 software, which requires the land use land cover map and applies specifi c equations for each metric based on the Fragstats (Spatial Pattern Analysis Program for Quantifying Landscape Structure) developed by McGarigal and Marks in 1995. First, the analysis was conducted considering diff erent size classes of forest fragments: less than 1 ha, 1 to 9.99 ha, 10 to 19.99 ha, 20 to 39.99 ha, 40 to 79.99 ha, and bigger than 80 ha. Subsequently, we conducted an individual analysis of the fragments bigger than 40 ha since they had great relevance in the land cover and higher stability due to the nuclear area. The edge eff ect considered for forest fragments was 62 meters, which was the average value of the interval (50-78 m) obtained by Lemos (2008) in a region near the studied basin.

Physiographic and pedological representativeness of land uses
To increase the details of our analysis, we generated a DEM (Digital Elevation Model) in the ArcGIS 10.1 Topo to Raster interpolator with 5 x 5 m resolution using the level curves (5 meters) and the mapped hydrography as inputs. The level curves, with class A altimetric accuracy, were obtained from the digital restitution of the Ikonos II image from October 2007. The hydrography was obtained by visual mapping of the Ikonos image, with the aid of the image available in Google Earth (Digital Globe, 2013), of the level curves and the hydrography mapped from IBGE. Field visits were performed with GPS to ensure the correct mapping of hydrography in places considered visually non-conclusive.
The analysis of representativeness of land use was analyzed by the compartmentalization of the relief in relation to geomorphology, exposure faces, and soils. For automatic segmentation of relief forms, we combined the attributes of the slope, curvature, accumulated surface fl ow and Euclidean distance to the waterways to identify four morphological units: hilltops, river plains, and terraces, concave and convex slopes. The criteria of automatic segmentation, adopted by Sarcinelli et al. (2012), had the limits of the variables adjusted according to the regional specifi cities of the relief of the research site.
The representativeness of the land use land cover in the various faces of exposure was analyzed from the exposure face map with results obtained in degrees with a specifi c notation for each class: fl at (< 3% slope); North (0 -45° and 315 -360°); East (45 -135°); South (135 -225°); and West (225 -315°).
For the analysis of pedological representativeness of soil uses, we used the soil map produced by Bastos et al. (2009), and the key to identifi cation of the classes was based on the soil/landscape model of the region as proposed by Corrêa (1984).

Statistical analysis
Statistical analysis of the representativeness was analyzed by identifying whether the forest fragments were well represented in relation to the attributes evaluated in each of the three compartmentalities. A contingency table was generated with the number of cells that each class of variables analyzed occupies in each of the land uses present in the basin. The relative deviations were analyzed concerning the expected values for the forest fragments in the compartmentalities performed. The Chi-Square test was performed to verify whether the observed absolute frequency of a variable diff ered signifi cantly from the expected absolute frequency distribution, according to Equation 1: Where, "o" represents the frequency observed for each class and "e" the expected frequency for each class.

Analysis of landscape fragmentation by size class
In general, natural forests presented fragments of high medium sizes and good proximity between the fragments. Approximately 33% of forest fragments have a dimension of less than 1 hectare (Table 1). On the other hand, the class of larger size (greater than 80 ha), has only one forest fragment (234.6 ha). In this fragment is located the Forest of Paraíso Table 1 -Class-level landscape structure for the size classes of forest fragments. Tabela 1 -Estrutura da paisagem em nível de classe para os as classes de tamanho dos fragmentos fl orestais. Research, Training and Environmental Education Station, which is responsible for an area of 194 ha of forest area. Although there are few large fragments, fragments with areas greater than 40 ha total cover approximately 51.6 % in relation to the total forest.
The average size of the fragments increased with the increase of the size classes, as well as the standard deviation of the sizes, however, the fragments smaller than 10 ha (Table 1) presented a high coeffi cient of variation. The shape of the fragments tended to be more irregular in the smallest fragments and with a lower proportion of the nuclear area.
The landscape metrics of the fi ve largest fragments (with areas greater than 40 ha) are presented in Table 2, and fragments between 45.2 and 234.6 ha were observed. The edge indexes of the fragments have higher values in fragments 3 and 5, and they also had the most irregular forms.
Fragment 3 had the lowest central area index (18.9%) and resulting in 5 distinct nuclear areas. In contrast, fragments 1, 2, and 4 presented similar central areas index and only one nuclear area for each. Fragment 5 presented the highest index of central area, with a value of 61.5%, distributed in 3 nuclear areas. The proximity index has led to the fragments being well connected, except for fragment 3, which has 572 m from the nearest fragment.

Representativeness of physiographic and pedological attributes
The spatial relationship between land use and topographic, pedological, and face compartmentalization (with 18, 12, and 24 degrees of freedom, respectively) was not signifi cant according to the values of Chi-square tests at 95% probability. The topographic compartmentalization (Table 3, Figure  2a) showed that there was a much lower occupation of forests (40.0%) in the areas of river terraces than expected. In the areas of convex slopes, there were also native forests with values below the expectations, with a deviation of 9.6%. The forest occupies more frequently than expected the areas of hilltops (28.3%) and concave slopes (8.8%).
The statistical analysis of pedological compartmentalization (Figure 2b) showed that forest has low representativeness in the areas of Red-Yellow Acrisols, with a value of 60.1% below the expected. In the Red-Yellow Ferralsols, the forest was represented with 42.2% of occurrence more than expected, and this result was positive in terms of soil water storage. In the Region of Haplic Cambisols, the forest representation value was 10.1% higher than expected ( Table 3). The forest presented higher values than expected on the south, east, and fl at faces (47.6, 10.1, and 4.8%, respectively) for the compartmentalization of the exposure faces, and values below the expected in the north and west faces, especially the north face with 28.5% less than expected (Table 3, Figure 2c).

Landscape structure
Most of the forest fragments present in the São Bartolomeu river basin do not have suffi cient size and/or shape to maintain a nuclear area, which may be insuffi cient to maintain the biodiversity of most plant and animal species. Each species has its particular characteristics in terms of area and habitat types for survival (Fahrig, 2003) associated with food, shelter, and reproduction. A forest fragment of 25ha is the minimum size suggested for the fl oristic stability in the Atlantic forest biome (Metzger, 1997). However, smaller fragments reduce the isolation of forest fragments in the landscape and help to maintain the regional biodiversity, functioning as ecological trampolines (Ribeiro et al., 2020). With the increase of fragments' area, expressed in size classes, they became rarer and thus representing a lower frequency class in the landscape. This indicates the occurrence of few source areas for the maintenance of populations established in smaller fragments. The area of a fragment is, in general, the most important parameter to explain species richness variations (Pirovani et al., 2014). To Ribeiro et al. (2020), these larger remnants are of great importance for the maintenance of biodiversity and should be the focus of conservation projects. Therefore, the large fragments of the São Bartolomeu river basin have great ecological importance, highlighting the forest fragment in the Mata do Paraíso.
Among the indexes analyzed, the edge densities had more discrepant values among the size classes, with the larger classes presenting lower edge eff ects and lower susceptibility to external disturbances. The impact of the border eff ect is related to the shape of the fragment, the larger the border being the more irregular the shape. The smaller the fragment or the longer the fragment, the stronger the edge eff ect, as the inner/margin ratio decreases. The excess of edges represents an ideal situation for the development of species of exotic grasses that successfully inhabit the environment (Salomão et al., 2018).
The forest fragments showed a tendency of shape irregularity as the size class increased, intensifying the edge eff ect, and this situation was also observed by Oliveira (2011). Moreira (2009), working in the area of the São Bartolomeu river basin, found through the circularity index (CI) that the forest fragments had a strong edge eff ect, with 66 of the 78 fragments presenting elongated shape with CI below 0.503.
Our study found only 59 nuclear areas considering the edge eff ect of 62 m. However, the smaller fragments are considered important due to the proximity to large fragments, fulfi lling the role of insulating reducers in fragmented landscapes, which can connect the fragments by corridors or serving as stopping sites of migrating organisms (Guariz and Guariz, 2020). According to Pirovani et al. (2014), more regular fragments minimize the ratio border/ area, and therefore, the central area is farther from the edges and isolated from external factors, infl uencing the quality of the structure of these ecosystems in the future.
Fragment 3 presented the most irregular shape and higher values of edge indexes among the largest fragments, mainly due to its very jagged perimeter. The central area index reinforced the fragility situation of fragment 3, resulting in a division of the fragment into 5 distinct nuclear areas. In addition, it presented more isolation in relation to the others. In general, a low degree of isolation of the native forest was obtained, and it can be translated into great possibilities of interaction between the fragments, through the dispersion of propagules and the fl ow of animals. According to Gustafson and Parker (1994), a reduction in isolation produces exponential increases in the index, and the increase in the size of these fragments produced a more modest linear increase in the index. Metzger (1997) mentions that the connectivity between fragments is an important physical factor related to variations in tree species richness. Isolated communities may exhibit very variable composition and slow recovery patterns, but connectivity between disturbed fragments has increased disturbance resistance (Schackelford et al., 2018).

Physiographic and pedological representation
The curvature of the terrain infl uences the erosive power of the water fl ow on the surface, and also the water dynamics through the soil profi le. We found that the forest occurs in concave areas more frequently than expected, being this result a good indicator for water conservation. In the concave areas occurs the convergence of the water fl ow, favoring erosive processes that result in grooves and can evolve in a gully. According to Chagas et al. (2013), soils tend to become saturated, and infi ltration occurs in the lower parts of the slope and the elevated concave positions (surface fl ow of convergent water). The concave surfaces are characterized by microclimatic, topographic, and solar radiation conditions and accumulation of water, sediments, minerals and organic matter may occur in this pedoform (Aumond and Maçaneiro, 2014). This situation favors the occurrence of species with higher nutritional requirements and water demand. In contrast, the convex slopes present divergent water surface fl ow and favor laminar erosion. The lower percentage of native forest (8.8%) in the convex slopes than the expected results in lower problems related to erosion and formation of the formation of gullies, but the restoration of these areas with natural vegetation can provide greater soil stability by reducing laminar erosion, especially in the high slope areas.
The terraces along the watercourses are associated with Acrisols and a low slope (0-8 %) and are preferred for occupation by residence and agriculture (Ippoliti et al., 2005). However, Souza et al. (2013) found that fl at areas presented less erosion, highest diversity indexes, and higher numbers of rare species. The region of terraces with Red-Yellow Acrisols present in the basin, associated with the low occurrence of forest, are considered priority areas in forest recovery planning. Increase forest in these areas can favor biodiversity, but also protect the watercourses and the quality of water resources.
The forest representativeness in the Ferralsols promotes positive results in terms of water storage in the soil since Ferralsols have a great water storage capacity due to their physical nature. Ferralsols have a soil physical structure with high porosity and provide optimal conditions for water infi ltration and percolation through the soil profi le (Brandão et al., 2006). The Red-Yellow Ferralsols were also found on the tops, wavy slopes, and inclined planes. Ippoliti et al. (2005), in a study in the municipality of Viçosa-MG, observed that the Red-Yellow Ferralsols of fl attened tops are generally deeper and developed, while in lower parts, especially in the lower third of the elevations, it can still be deep in convex-convex pedoforms or be shallow in convexconcave pedoforms. In the tops of hills, the high forest representativeness can be associated with the Ferralsols of the fl attened tops, being important for its high water storage capacity in the soil profi le.
The Haplic Cambisols occupy the steepest slopes of the landscape, being shallow soils and more susceptible to erosion, often with exposure of the C horizon. Rezende (1971), in studies in the Zona da Mata Mineira, highlighted that Cambisols located in the very steepest areas are the least thick of the sequence, because this position favors morphogenesis, leading these soils to a constant rejuvenation by erosion. Our study shows that the forest presence in these areas was satisfactory and since the water surface fl ow is favored and can trigger erosive processes and cause the reduction of water quality, the proper soil use and management in these areas are extremely important.
In the Southern Hemisphere, the sun has an apparent east-west trajectory with declination to the north, and it results in solar light tending to intercept the surfaces facing north and tangentially those facing south (Ferreira et al., 2005). According to this author, this eff ect will be more pronounced in areas with wavy relief. Knowing that the angle of incidence of the sun's light promotes changes in the amount of heat transmitted (Bonan, 2002), the surfaces facing north receive more energy than those facing the south, tending to have higher potential evapotranspiration and lower soil moisture.
In this study, the forest presented the lowest representativeness on the north face, which may represent losses of forest species that require a higher incidence of solar radiation and adapted to drier soils. Due to the lower protection of areas with lower water availability, populations of species not tolerant to fl ooding or the high water table may be underrepresented (Sarcinelli et al., 2012).
The east and west faces presented diff erences related to the time of solar radiation incidence since the total amount of incidence time is the same. The eastern face receives radiation in the morning, a period from thermal inertia from the night before, presenting lower temperatures and favor the maintenance of soil moisture. Because the natural vegetation is distributed predominantly on the fl at areas, eastern and southern faces, good representativeness of forest species adapted to humid and more shaded environments is expected.
In general, the places with a higher incidence of solar radiation, mainly located in more fertile soils associated with convergent water fl ow regions and deposition areas, had low natural vegetation, being therefore considered niches of rare presence in the landscape associated with fauna and fl ora. Batista et al. (2009) highlight that studies based only on the most preserved fragment(s), as commonly occurs in management plans, do not guarantee the conservation of the natural diversity of land and the vegetation gradient that determine local biodiversity.
Production of water associated with biodiversity conservation is one of the main ecosystem services of the São Bartolomeu river basin. Then, it is necessary to ensure forest representativeness in all types of environments of the basin, favoring the sustainability and resilience of natural vegetation associated with good hydrological conditions, ensuring the interaction of processes between forest fragments and, consequently, of the landscape as a whole. Native forests promote the production and conservation of water sources, favoring the reduction of peak runoff , distributing their occurrence in a longer period of the year with good quality water. Silva et al. (2020) corroborate that actions are needed aiming to recover the recharge potential in the basin, especially with a focus on conservation practices that foster soil protection and appropriate land use land cover in the contribution area of the São Bartolomeu river basin.

CONCLUSIONS
The forest fragments present good proximity, however, they have irregular shapes that favor the edge eff ect. The existence of few source areas for the maintenance of populations established in smaller fragments is diff erent from the importance of connectivity between the fragments. Therefore, the management in this basin should take these fragments into account as a way to connect with large fragments and improve regional biodiversity. Although the study of landscape structure is of fundamental importance, habitat divisions in the fragmentation process may not be suffi cient to understand the eff ects on biodiversity, as it does not consider the diversity of niches and landscape compartments.
The largest forest fragments present the highest rates of nuclear area, showing that these fragments are large enough to minimize the eff ects of edge and generate fragments with large nuclear areas. Therefore, these large fragments must be conserved in management planning aiming at water conservation and biodiversity protection, Among the fi ve largest fragments, a fragment showed results that indicate the need for practices that improve its stability, such as forest restoration and land uses in the surroundings that decrease disturbances and other conservation measures.
The distribution of forest fragments in specifi c niches in relation to soil type, facing slopes and geomorphology should be associated with studies that support the decision about priority areas for forest conservation. In general, our study shows that there is a compartmentalization that may not guarantee the maintenance of biodiversity. In order to increase the spatial representativeness of the native forest in the landscape, forest use should be a priority in the terraces (riparian forests) nearby the rivers and associated with the Red-Yellow Acrisols, convex slopes, and on the north and west faces. However, there must be a balance in the diff erent compartments, associating with forest connectivity and landscape structure.