Vector–raster server-side analysis: a PostGIS benchmark

Abstract

Report generation from coverages around points of interest (POIs) or in areas of interest (AOIs) is a common need in thematic research projects. The extracted information adds value to the POIs or AOIs dataset by enriching its information content. In a common scenario, POIs or AOIs are usually vector data whereas background thematic datasets are often raster data. The paper investigates different approaches for zonal statistic computation about both raster and vector data with particular focus on server-side free and open source software (FOSS)-based solutions. Extensive performance tests are based on PostGIS (the spatial extension of popular PostgreSQL (http://www.postgresql.org/) FOSS DBMS) 2.0. This version is the first to offer raster support. Previously, vector–raster analysis was not supported by any FOSS DBMS environment and such analyses were possible in a FOSS server-side environment using geographical information system (GIS) tools (e.g., Geographic Resources Analysis Support System (GRASS) GIS) and Open Geospatial Consortium (OGC) Web Processing Server. PostGIS performance data from the tests are compared to an almost standard ESRI ArcGIS desktop approach. A project aimed at monitoring fire alerts in African protected areas provides the benchmarking application. The computation of people living in the surroundings of POIs (alert points from Global Fire Information Management System) based on a world population density dataset (LandScan) is the benchmarking query. The impact of many parameters on performances is considered: the adopted tile size in the storing of the raster in the DBMS, the dimension of queried areas in relation to the above-mentioned tile size, the number of queried features.