The pervasive usage of geoprocessing tools between researchers, students and users’ communities leads to resources integration and allowed online tools to accomplish demands of people’s daily life. With the progress of Information and Communications Technology (ICT), new computational resources are constantly becoming available and the effort of research institutes goes towards to release online interactive systems to the public that allow geospatial data visualization and manipulation. However, behind all provided resources there are a lot of process dealing with data preparation, storage and recovery. In this way, this work was carried out to deal with these tasks, implementing logical models of geographic database in order to store land use and land cover maps produced for the Brazilian Legal Amazon. The dataset used to handle the experiments was obtained from TerraClass initiative, a project created to qualify the deforested areas for that region. The main objective of this study was to identify which thematic classes intersect with a set of municipalities in the state of Pará, returning results in an acceptable time window for a web application. To accomplish that, a PostGIS spatial database was established and geoprocessing techniques and spatial operators were used to perform the tasks. To speed-up the queries execution time, a new model was proposed, reorganizing data in a new object structure which results in an 80 times faster performance of data queries. Thereafter, indexes were implemented and the cost of query execution was optimized more than 50%, addressing an important issue regarding a multi-user environment scenario.