The Amazon Rainforest is the scene of a large number of deforestation activities such as artisanal mining, agriculture and timber trade. For the purpose of have reduced environmental impacts, regulatory agencies have attempted to regulate these activities and direct them towards more responsible methods of operation. This paper describes the initiative by monitoring these forest areas located near these regions of deforestation, because the core elements, such as biomass and carbon accumulation of the trees can be adequately monitored against occasional disturbances brought by these activities. The current standard approach in the Amazon is to monitor all the trees of the forest within an area called the transect, also designated as forest inventory, keeping a strict record of their behavior and growth. However, these activities are restricted to control areas that are located in strategic regions and do not represent the whole area to be monitored. This research explores a new methodology based on geostatistics and designed to optimize the sampling, extending the study of much larger forest areas, keeping unchanged the use of human resources unit, and at the same time increase the surface areas of study and to maintain confidence in the results. The proposed methodology allows the selection of the Legal Reserve-RL, to be made according to the actual distribution of carbon accumulation in the forest, instead relying in using area percentage proposed by law and common sense of proprietary / regulatory agencies. This methodology was applied in the Tapajós National Forest (FLONA Tapajós), State of Pará, Brazil, we used the data set available, to optimize the sample and monitor the forest's ability to store carbon. This methodology intends to contribute to reducing the cost of monitoring activities per unit area, increased precision for location RL, and simplifying procedures by applying a set of easy to use tools. The results showed that application of geostatistical studies for determination of RL is a viable procedure, because the structure of the variogram is maintained even with a random sampling suffering decreased to 50% of the area of vegetation, even managing to keep the sampling result the total vegetation cover.