The advent of advanced graphics technologies and improved hardware has enabled the generation of highly complex models with huge number of triangles. Point-based representations and rendering have emerged as viable representations for high-quality models in this scenario. These methods have been demonstrated only on high resolution, compact objects so far. They have to be adapted to environments that extend over large areas to be considered serious representations. In this paper, we present an adaptation of the point-based representation to large, hierarchical environments. We show how point-based data can be generated by sampling polygon-based representations. We also show the combination of an object hierarchy and multiresolution point representations and develop rendering algorithms for the same. The multiresolution representation is constructed during the generation process. We then show a hybrid representation in which the more complex portions of the environment are represented using points and others using the original polygon-based representation. This produces better rendering performance by keeping large, flat regions as triangles. We demonstrate the method on the model of the Fatehpur Sikri which has 14000 objects with over 500,000 triangles.