GDMS-R: A mixed SQL to manage raster and vector data

11pInternational audienceTo evaluate urbanization impact on territories, an accurate knowledge of the urban and peri-urban fabrics is unavoidable. To provide advanced characterization of the terrain, modern GIS applications target even wider geographic areas at finer resolutions but they also have to mix data of different types such as Digital Elevation Model (raster layer), buildings (polygonal layer) and roads (polylines layer). Processing both raster and vector data with the same semantic and in an efficient way presents significant challenges to GIS insofar as underlying granularities but also data layout and processing patterns might be absolutely different. We have already focused on the definition and the implementation of an abstraction layer called GDMS (Generic Datasource Management System) to handle and process vector data. Main objectives with GDMS, were to provide the user not only a simple and powerful API but also a spatial SQL derived language. Moreover, as an intermediate layer between the user and the information source, GDMS intends to reduce the coupling between the processes and the specificities of each underlying format. As a consequence, former work may easily be reused in a much larger set of scenarii. The learning curve is consequently even simpler. In this paper, we propose a raster extension to the GDMS layer called GDMS-R. Even if, there is currently no OGC standard concerning raster processing (using well-known SQL language), there already exists a de facto standard called Map Algebra defined by C. D. Tomlin in 1990 and commonly implemented in a wide set of GIS. Our objective is a bit different insofar as we propose to extend SQL language. We present the integration of Map Algebra concepts in GDMS through the GRAP (GeoRAster Processing) language. As for GDMS, reuse is enhanced by the possibility of being vendor-independent (middle-ware approach) and the extension capabilities of the underlying SQL language. To demonstrate the capabilities of GDMS-R, we present a use case relative to the deep impact of increased urbanization on the vulnerability of peri-urban hydro-systems: impact of the linear constraints on the runoff water pathways and accumulation that uses both vector and raster data in an unified way