Modelagem ambiental de espécies de árvores no Vale do Jari, Monte Dourado, Pará usando dados de inventário florestal.

Abstract

(Environmental Modelling of tree species in the Jari Valley, Monte Dourado, Pará state using a commercial forest inventory). Data on forest inventories are potentially unknown resources for use in environmental modeling. In contrast to herbarium data, they deal with a great number of individuals with a reasonably well-located geographic location and are available for much larger areas than those of permanent plots, (Hubble type plots). Problems in the use of these data include trustworthy identification of the species and the non-continuous cover of the areas. The present study was carried out in the forest management area of the company Orsa Florestal Ltda in Monte Dourado Pará. We used the area subdivisions as defined by the company for the forest inventory. These consist of Annual Production Units (UPA in Portuguese) with irregular shapes and sizes, subdivided into 10-ha Work Units (400m x 250m). The tree data bases are presented in a table with the list of georeferenced species from the forest inventories and the environmental-variable data bases contain relief curves, hydrology and topographic classification given by the company. Data from 4 UPAs were used; data from UPAs 1 and 2 were used for studies of floristic composition, diversity, spacial structure and in modeling for prediction of distribution; UPAs 3 and 4 were used only in species distribution modeling . Ten key species were selected from the species list for modeling and spacial distribution characterization; these were defined as those that presented the highest values of abundance and dominance. Key species identification was checked by collectiing botanical material and comparing it with herbarium material (IAN, Eastern Amazônia Embrapa herbarium).. The precision of the georeference data on the trees was also confirmed in field. Information on relief curves (altitude and declivity), hydrology and topography supplied by the company were analysed with ArcView 9.2 software to generate the tables to be used to model geographic distributions of the key species. Composition and plant physiogonmies were defined and characterized, as well as diversity patterns and species richness was estimated for the study area. The key species were studied as regards spatial and population structure. Maxent and GARP were used to produce distribution models for these species. Nine topographic sites were selected on the basis of the environmental variables, presenting high diversity (Hm'=3.85±0.27). Sites within the same altitude and declivity class are more closely grouped in lowland environments, while for hillside and plateau sites, the variable distance from the river confers greater affinity between sites. Vouacapoua americana had the highest dominance values in most sites, predominating hillside and plateau environments. In lowland environments, the highest dominance recorded was for Goupia glabra and Manilkara bidentata ssp. surinamensis. Diversity was influenced exclusively by the variable distance from the river, the greatest diversity being found in periodically flooded habitats. Dominance and abundance were influenced by the association altitude/declivity and by distance from the river; the highest dominance domiance values were found at higher areas (hillsides and plateaus) and nearer the river. Highest abundance values were found in lowland environments and at intermediate distance from the river. The richness estimators used showed average representation of 88% in relation to the observed data. A clumped pattern of spatial distribution did not prevail in the plots. Effect of the altitude/declivity variables showed a trendency for regular patterns. The algorithms MaxEnt and GARP were not efficient in predicting species occurrence in areas with different topographical characteristics, even for species with abundance values independent of these factors (e.g. G. glabra, D.excelsa, M.huberi). However, good models were obtained, when considering the entire topographic variation of all environments, and also when the number of points used to generate the model was less than those used for validation. MaxEnt showed the best power of prediction of the models; however, extrapolação of prediction to new areas should consider species abundance and include more environmental variables. Forest inventories supply data capable of demonstrating richness for studies of diversity and spatial distribution of species in Amazonian forest systems and, therefore are valuable tools for ecological studies in the Amazon.