LandEx – A GeoWeb-based Tool for Exploration of Patterns in Raster Maps

Advances in remote sensing and GIS make possible construction of high resolution categorical raster-based maps depicting spatial distribution of natural and/or anthropogenic features. The best know example is a land cover/land use (LCLU) map. GIS tools for performing queries for spatial extent of single map category are readily available. However, often, an analyst would like to query a map for an occurrence of a particular spatial pattern of categories rather than a single category; no tools exist to execute such query. We have developed the Landscape Explorer (LandEx) – an algorithm that performs query by pattern similarity (QBPS) on categorical rasters – to provide such functionality. In the context of land cover, a pattern of categories may be associated with semantic notions of “downtown”, “typical small town”, or “an irrigated agriculture in a desert environment” to give just a few examples. QBPS finds regions across the map having patterns of categories (and thus, presumably, semantic meanings) similar to a pattern of a reference region. Fig.1 illustrates an idea of LandEx in a nutshell using the National Land Cover Dataset 2006 (NLCD2006) (Fry et al., 2011) map as an example. In this example LandEx takes as input a categorical raster map with K=16 land cover classes. A user selects a small region of interest (ROI) and wants to identify other regions across the conterminous United States having similar patterns (and thus similar semantic meanings). LandEx outputs map of similarity values thus enabling a user to put the ROI in a broader geospatial context. Development of LandEx involves two separate tracks, one is a design of the core algorithm that executes a QBPS request and another is an implementation of this algorithm as a GeoWeb service. Our goal for LandEx is to offer a tool capable of intelligent knowledge discovery in large (or very large) spatial datasets in a convenient and intuitive manner, but also capable of delivering research-grade results. To fulfil this goal LandEx is accessible at http://sil.uc.edu/landex as a GeoWeb service that provides all expected functionalities.