GeoPAT 2.0 - Software for Pattern-Based Spatial and Temporal Analysis of Large Earth Science Datasets
Presenting Author: Tomasz Stepinski
Organization: University of Cincinnati
Co-Author(s): Pawel Netzel, Jacek Niesterowicz
GeoPAT 2.0 (Geospatial Pattern Analysis Toolbox) is a standalone suite of modules written in C and dedicated to analysis of large Earth Science datasets using spatial or temporal patterns. It is a new code based on the idea of an original GeoPAT - a suite of GRASS GIS modules working under Linux. In addition to being freed from GRASS GIS and dependence on Linux, GeoPAT 2.0 extends the domain of applicability from categorical rasters and spatial analysis alone to wide variety of data formats and spatial as well as temporal analysis. Full parallelized GeoPAT 2.0 code for Xeon Phi will also be available. GeoPAT 2.0 works with patterns. It transforms large gridded datasets such as, for example, the global 300 m resolution CCI land cover or the global 90 m resolution SRTM topography into grids of motifels – units of pattern of original variable. All subsequent analysis is performed on motifels using an appropriate measure of dissimilarity between them. GeoPAT has modules for similarity query, grid segmentation, and change detection. It contains modules implementing various methods for pattern representation as well as modules implementing several pertinent pattern dissimilarity measures. It has been tested on global and continental-scale land cover datasets, on global topography dataset, on global climate grid dataset, and on very high resolution images. GeoPAT provides computational engine for several Web-based applications performing query-by-example exploration of large land cover, topography, and climate datasets. These applications are available at http://sil.uc.edu .