Floristic surveys allow us to make a momentary assessment of the structure and degree of conservation of the vegetation, supporting future actions for the sustainable use of forest resources. The aim of this work was to study the floristic and structural composition of fragments of floodplain forest in the Cajari River Extractive Reserve, Amapá state, Brazil. Systematic sampling was applied with the allocation of 27 plots of dimensions 10 m x 100 m each, with inclusion level of individuals greater than or equal to 10 cm in diameter at chest height. The floristic diversity of species was determined by the Shannon Diversity Indexes and Pielou Equability. The characterization of the horizontal structure was determined by the parameters: density, frequency and relative and absolute dominance, values of ecological importance and the diametrical distribution. The families Fabaceae and Euphorbiaceae were the ones that obtained the highest richness of species and number of individuals. The species Virola surinamensis Warb., Hevea brasiliensis (HBK) Muell. Arg., Pentaclethra macroloba (Willd.) Kuntze and Mora paraensis (Ducke) Ducke, were the most abundant. The várzea forest fragments presented high diversity and high ecological dominance of species. The J-inverted diametric distribution, characteristic of natural forests, indicates that it is a mature forest component with an expressive group of dominating (hyperdominant) species, such as the species V. surinamensis and M. paraensis. Most species presented aggregate distribution. It is hoped that these results may serve the environmental control and control bodies as strategies for the use of forest resources and in the design of the management plan of Resex.

Accurate quantification of above-ground biomass (AGB) in managed forests requires: consideration of inventory errors and the use of local or large-scale allometric models. In this study we focus on the measurement errors, data collection errors and we compared different methods to estimate AGB in managed tropical forest. The data were collected in 15 plots of 100 x 100 m. We evaluated the errors of the forest inventory of 8.898 trees. We used four methods to estimate AGB: three methods which use a pan-tropical equation, which depends on wood density data, with different ways of integrating the wood density data (obtained from dataset of the Brazilian Forest Service, Jari and Global Wood Density Database - GWDD); and one local equation. The main inventory errors were: problems with the same tree being identified as a different tree in consecutive measurements (16% of the trees). AGB estimates using each of the four methods were significantly different.