Caracterização e perspectivas do Parque Estadual do Rio Doce - MG: uma abordagem a partir de imagens de sensoriamento remoto e fotografias hemisféricas de dossel

Abstract

The Atlantic Forest, an important biome characterized by high biodiversity and endemism, is formed by a group of forest ecosystems forests and associated ecosystems. Despite its magnitude, this hotspot is extremely threatened, since it is estimated that 93% of its original formation has been devastated. Deforestation and conversion of primitive forest in other types of cover are still reducing the native forest and is at the small percentage of the Atlantic Forest that remains in Minas Gerais which the Rio Doce State Park (PERD) is located, considered the largest continuous patch of this biome in the state, containing many of the lakes of the middle Rio Doce lake system, the third largest in the country. Forest ecosystems with high degrees of fragmentation and isolation of remnants, such as the Atlantic Forest of the middle Rio Doce valley, are especially susceptible to a severe process of biodiversity reduction and the high levels of endemism turns it even worse. The land cover and uses classification of PERD and surrounding from high-resolution satellite images, like IKONOS, using manual methods, expands the understanding of the landscape matrix in which the Park is located. This type of image provides high accuracy and high detail levels, what reflects into the produced thematic maps quality. In the tropics, the forest is a mosaic of successional stages with imprecise limits and it may require a more refined vegetation cover classification. The canopy hemispherical photograph technique, that uses wide-angle fisheye lens, has been used in many canopy structure and forest light transmission studies. The classification of this type of image, through GLA software, allows the differentiation between forest vegetation classes, including successional stages of Atlantic Forest. The extracted products may be used as auxiliary information in the differentiation of forest types and from them was possible to propose a preliminary model of forest vegetation cover, which differentiates native forest in different successional stages, eucalyptus plantation and mixed areas. Although it is preliminary, the proposed model seems to be adequate for the differentiation of the considered classes. A sample increase and an adequacy of technique use will allow improvement of the model. The observation prospects approached are pioneer initiatives. The results achieved in this research provide insights and contributions that will enable the targeting of research efforts and government actions, both important to Atlantic Forest management and conservation