IN22A-03:
Analysis of Giga-size Earth Observation Data in Open Source GRASS GIS 7 - from Desktop to On-line Solutions.

Tuesday, 16 December 2014: 10:50 AM
Helena Mitasova, North Carolina State University at Raleigh, Raleigh, NC, United States, Tomasz F. Stepinski, University of Cincinnati Main Campus, Space Informatics Lab - SIL, Cincinnati, OH, United States, Jaroslaw Jasiewicz, Adam Mickiewicz University, Geoecology and Geoinformation Institute, Poznań, Poland, Markus Neteler, Fondazione Edmund Mach and Innovation Center, San Michele all'Adige, Italy and Soeren Gebbert, Thuenen Institute of Climate-Smart Agriculture, Braunschweig, Germany
Abstract:
GRASS GIS is a leading open source GIS for geospatial analysis and modeling. In addition to being utilized as a desktop GIS it also serves as a processing engine for high performance geospatial computing for applications in diverse disciplines. The newly released GRASS GIS 7 supports big data analysis including temporal framework, image segmentation, watershed analysis, synchronized 2D/3D animations and many others. This presentation will focus on new GRASS GIS 7-powered tools for geoprocessing giga-size earth observation (EO) data using spatial pattern analysis. Pattern-based analysis connects to human visual perception of space as well as makes geoprocessing of giga-size EO data possible in an efficient and robust manner. GeoPAT is a collection of GRASS GIS 7 modules that fully integrates procedures for pattern representation of EO data and patterns similarity calculations with standard GIS tasks of mapping, maps overlay, segmentation, classification(Fig 1a), change detections etc. GeoPAT works very well on a desktop but it also underpins several GeoWeb applications (http://sil.uc.edu/ ) which allow users to do analysis on selected EO datasets without the need to download them. The GRASS GIS 7 temporal framework and high resolution visualizations will be illustrated using time series of giga-size, lidar-based digital elevation models representing the dynamics of North Carolina barrier islands over the past 15 years. The temporal framework supports efficient raster and vector data series analysis and simplifies data input for visual analysis of dynamic landscapes (Fig. 1b) allowing users to rapidly identify vulnerable locations, changes in built environment and eroding coastlines. Numerous improvements in GRASS GIS 7 were implemented to support terabyte size data processing for reconstruction of MODIS land surface temperature (LST) at 250m resolution using multiple regressions and PCA (Fig. 1c) . The new MODIS LST series (http://gis.cri.fmach.it/eurolst/) includes 4 maps per day since year 2000, provide improved data for the epidemiological predictions, viticulture, assessment of urban heat islands and numerous other applications. The presentation will conclude with outline of future development for big data interfaces to further enhance the web-based GRASS GIS data analysis.