The novelty of this scenario is to convert terrain data to a multi-resolution representation suited for analysis. The terrain model is based on a triangle mesh, and organized in a multi-resolution structure that reflects the hierarchy of drainage basins. We re-organize the high-resolution elevation model, given as LAS point clouds, according to semantic criteria such as the membership to a catchment basin. For environmental aspects, we arrange points according the membership of the points to watershed. Each watershed is grouped in a macro-set according two different solutions: the Catchment Area Plan, which reflects homogeneous regions with respect to environmental legislation, or the Warning Areas, which groups together area with similar meteo-hydrologically response. The catchment boundaries are available from the related cartography owned and maintained by the Civil Protection group.

The workflow is divided in two parts.

1. The first part  is a pre-processing step, which is run only once when new data is added and creates a multi-resolution indexing of the input point cloud.
2. The second part LS1.2 queries the output of LS1.1 and produces a multi-resolution triangulation of a given set of regions that are selected by the user together with a pre-defined level of detail, see Figure 4.

The Regione Liguria LiDAR data set has 20 billion points (size: ~600TB), contained in ~1000 partially overlapped LAS files. The multi-resolution index is built in several phases: filtering, decimation, indexing according to basins, computation of multiple levels of detail. Computation of the index and computation of MLODs achieve parallelization using a scatter paradigm. The workflow currently runs in the cloud in ~2h using 12 nodes.

The figure bellow shows an example of a Multi-resolution representation of a drainage basin