Marcela de Castro Nunes Santos Terra

Assessment of geostatistical features for object-based image classification of contrasted landscape vegetation cover

Accurate mapping and monitoring of savanna and semiarid woodland biomes are needed to support the selection of areas of conservation, to provide sustainable land use, and to improve the understanding of vegetation. The potential of geostatistical features, derived from medium spatial resolution satellite imagery, to characterize contrasted landscape vegetation cover and improve object-based image classification is studied. The study site in Brazil includes cerrado sensu stricto, deciduous forest, and palm swamp vegetation cover. Sentinel 2 and Landsat 8 images were acquired and divided into objects, for each of which a semivariogram was calculated using near-infrared (NIR) and normalized difference vegetation index (NDVI) to extract the set of geostatistical features. The features selected by principal component analysis were used as input data to train a random forest algorithm. Tests were conducted, combining spectral and geostatistical features. Change detection evaluation was performed using a confusion matrix and its accuracies. The semivariogram curves were efficient to characterize spatial heterogeneity, with similar results using NIR and NDVI from Sentinel 2 and Landsat 8. Accuracy was significantly greater when combining geostatistical features with spectral data, suggesting that this method can improve image classification results.Accurate mapping and monitoring of savanna and semiarid woodland biomes are needed to support the selection of areas of conservation, to provide sustainable land use, and to improve the understanding of vegetation. The potential of geostatistical features, derived from medium spatial resolution satellite imagery, to characterize contrasted landscape vegetation cover and improve object-based image classification is studied. The study site in Brazil includes cerrado sensu stricto, deciduous forest, and palm swamp vegetation cover. Sentinel 2 and Landsat 8 images were acquired and divided into objects, for each of which a semivariogram was calculated using near-infrared (NIR) and normalized difference vegetation index (NDVI) to extract the set of geostatistical features. The features selected by principal component analysis were used as input data to train a random forest algorithm. Tests were conducted, combining spectral and geostatistical features. Change detection evaluation was performed using a confusion matrix and its accuracies. The semivariogram curves were efficient to characterize spatial heterogeneity, with similar results using NIR and NDVI from Sentinel 2 and Landsat 8. Accuracy was significantly greater when combining geostatistical features with spectral data, suggesting that this method can improve image classification results.

Relação Espacial do Carbono da Vegetação e Matéria Orgânica do Solo na Serra da Mantiqueira

We aimed to analyze the spatial behavior of carbon present in the stems of trees (ECV) and soil organic matter (OM) in a dense ombrophilous forest at Serra da Mantiqueira, MG, using kriging. Also we tested the use of superficial organic matter as auxiliary variable in predicting ECV by cokriging. In order to achieve this, we analyzed georeferenced data from 25 sampling points of soil organic matter in three depths, 0-20 cm (P1), 20-50 cm (P2) and 50-100 cm (P3) and 12 vegetation sample plots of 400 m2. All variables presented spatial dependence structure (higher ECV and lower OM at P1). The cross semivariogram reflected the spatial correlation between ECV and OM at P1. The interpolation map generated by cokriging provided good notion of general ECV trends keeping the same patterns as the one generated by ordinary kriging for this variable. Our findings pointed that in our studied area, OM at P1 can be used as a covariate for ECV prediction of where ECV is under sampled.

What can natural edges of gallery forests teach us about woody community performance in sharp ecotones

Aims The diversity and structure of forest edges are a central issue in ecology. However, most studies have focused on anthropogenic forest edges, being natural ones least understood. We studied the communities of shrubs and trees in natural edges of gallery forests with the main goal of learning what are the main factors that shape the structure and diversity of these natural sharp ecotones. Methods We evaluated 10 gallery forest sites, allocating in each of them three 15 × 20 m plots. The plots were laid out in relatively rectilinear stretches of forest edges, respecting a minimal distance of 10 m between each plot. As they are permanent plots and meant to study eventual fluctuations in the forest–grasslands limits, we allocated the plots with their longest side parallel to the forest edge and covering perpendicularly 5 m of the grassland and 10 m of forest. Inside the plots we identified, mapped and measured all shrub and tree individuals with diameter at breast height (DBH) ≥ 1 cm. Important Findings Although many differences exist among the communities in terms of density and basal area, all of them were highly dense and had relatively low basal areas. In addition, both the number of individuals and the basal area increased rapidly towards the forest interior. Our results also revealed that the edge communities are composed mainly by small individuals that demand a great amount of light. Both the vertical stratification and the proportion of the functional groups are similar among the sites yet change quickly in the first 10 m of the forest edge. However, the floristic similarities were low among the communities mainly due to the species turnover within each area. Therefore, although the overall structure is constant in the studied edges, the low floristic similarity among them possibly indicates that the species within the same functional groups are substituting each other across the sites. Despite older and seemingly stable, our results showed the gallery forest edges are quite similar to the anthropogenic edges or recent clearings. This points out that, independent of the age, those characteristics are possibly permanent in edges.

SHORT-TERM CHANGES IN AN OMBROPHILOUS ATLANTIC FOREST

Characterizing and monitoring the Brazilian Atlantic Domain provide important direction for the conservation of its remnants. Short-term variations in forest composition and structure are usually lost in larger intervals between inventories and could better elucidate forest dynamics. This study aimed to analyze temporal changes in the structure, floristic composition and diversity of a tree community of an Ombrophilous Dense Forest remnant in Bocaina de Minas, Minas Gerais state, Brazil, in two short intervals. The study was conducted based on circumference at breast height (1.3 m) (CBH) data from three forest inventories (2009, 2011, and 2012) in twelve 400 m2 permanent plots randomly distributed in the area. The results point out to a forest characterized by the increase in plant density and basal area, mainly due to the continuous local anthropogenic disturbances observed in the area, such as the presence of cattle and trails. These disturbances start early successional processes in various sectors of the forest. The altitudinal gradient leads to great environmental heterogeneity, resulting in a high species turnover, as well as high values of diversity which remained unchanged in the intervals considered.

The influence of disturbance on driving carbon stocks and tree dynamics of riparian forests in Cerrado

Aims Riparian forests in the Brazilian Cerrado, also known as gallery forests, are very heterogeneous in structure, species composition and ecological features due to strong and abrupt variations of soil, hydrological and topographic properties. However, what are the variables driving forest carbon stock and productivity, mortality and recruitment in disturbed gallery forests? Methods We used 36 permanent plots data from a gallery forest in the Brazilian Cerrado. We investigated how tree community dynamics vary in a gallery forest on two contrasting disturbance levels—logged and non‐logged—across a topographic gradient intrinsically related to differences in moisture conditions, edge effects, as well as soil fertility and texture. Important Findings Soil variables were reduced into principal components and we used structural equation modelling to disentangle covarying variables. We also included carbon stocks as a determinant variable of dynamics rates. Logged forest had 50% higher productivity than non‐logged forest and streamside forest had aboveground carbon stocks 70% higher than the forest edge. Both logging and natural disturbance drove variation in the carbon stocks which contributed to shaping productivity and recruitment rates. Distance from the river also drove mortality and carbon stock rates. Areas with high‐carbon stocks favoured higher competition and lessened productivity and recruitment rates. Although soil fertility and texture are considered crucial components shaping forest dynamics, there was no clear influence of those variables on the present forest, probably because the strong effects of soil moisture, forest edge and disturbance disrupted the correlation between soil and forest dynamics.

Water Availability Drives Gradients of Tree Diversity, Structure and Functional Traits in the Atlantic-Cerrado-Caatinga Transition, Brazil

Aims Climate and soil are among the most important factors determining variation in tree communities, but their effects have not been thoroughly elucidated to date for many vegetation features. In this study, we evaluate how climate and soil gradients affect gradients of vegetation composition, species diversity and dominance, structure and functional traits (seed mass and wood density) using over 327 000 trees in 158 sites distributed along environmental gradients in the transitions among the Atlantic forest, Cerrado and Caatinga in Minas Gerais State (MG), Brazil (nearly 600 000 km 2). Methods Gradients in species, genus and family abundance in addition to basal area, stem density, species diversity (Fishers alpha), dominance percentage, seed mass and wood density were correlated using multiple regressions with environmental variables, as summarized in four principal component analysis axes (two climatic - precipitation seasonality and temperature range - and two edaphic - soil fertility and soil moisture). Additionally, ordinary kriging maps were used to better illustrate the gradients. Important Findings Multiple regression models indicate that all variables but dominance percentage were affected by one or more of the environmental gradients, but the average R 2 was low (26.25%). Kriging maps reinforced the patterns observed in the regression models. Precipitation seasonality and soil moisture gradients were the most important gradients affecting vegetation features. This finding suggests that water availability is an important determinant of vegetation features in these vegetation transitions.

Tree Responses to Soil and Edge Effects in a Semideciduous Forest Remnant

The present study seeks to characterize the composition, structure, and functional traits of the tree stratum of a Semideciduous Forest in Pouso Alegre, MG, Brazil (22o12’43”S and 45o59’30”W), by comparing Edge and Interior habitats and their relation to human impacts and soil variables. The tree community (diameter at 1.3 m (DBH) higher than 5 cm) was inventoried using the point-centered quarter method by laying out 12 transects of 450 m each (six transects in the Edge and six in the Interior). Functional traits – wood density and seed mass – were obtained by genus. The impact evaluation and soil sample collection were realized per transect. We found significant differences in terms of the basal area (higher in Interior), diversity (higher in Interior) and community seed mass (higher in Edge). The effects of higher values of soil traits such as Calcium, Magnesium, cation sum (CS) and base saturation (V) on species composition were stronger in the Interior, while human induced impacts and soil Potassium had an overwhelming effect on the Edge.

Temporal vegetation changes in a seasonally dry tropical forest enclave in an ecotonal region between savanna and semiarid zones of Brazil

Seasonally dry tropical forests (SDTFs) comprise a globally significant biome for biodiversity and conservation. Geographically, Brazilian SDTFs are primarily located within the country’s semiarid region (the Caatinga domain) in north-eastern Brazil. However, poorly studied and disjunct SDTF enclaves can occur within other regions, inside other Brazilian phytogeographical domains of vegetation such as savannas (i.e. the Cerrado domain) in central Brazil. These enclaves provide an opportunity to examine natural and non-anthropogenic edge effects on such vegetation. In 2007 and 2014, we studied a 120-ha SDTF enclave in the municipality of Januaria in northern Minas Gerais, Brazil, to understand its (1) floristic composition and soil correlates, and (2) temporal variations in diversity, structure and dynamics. Three sets of 10 400-m2 plots were used to compare the vegetation at 0 m (edge), 100 m (middle) and 200 m (inner) into the forest. The edge plots were compositionally dissimilar from the interior plots because of soil fertility and soil textural gradients. Paradoxically also, the inner plots exhibited less stable vegetation-dynamic patterns than did both the middle and the edge plots, possibly owing to natural temporal fluctuations in vegetation dynamics. Overall, the SDTF enclave exhibited high diversity and structural complexity, likely because of its geographical setting within a matrix of savanna. These results highlight a conservation priority for further studies on such SDTF enclaves throughout their range.