An information model for managing multi-dimensional gridded data in a GIS

Abstract

Earth observation agencies like NASA and NOAA produce huge volumes of historical, near real-time, and forecasting data representing terrestrial, atmospheric, and oceanic phenomena. The data drives climatological and meteorological studies, and underpins operations ranging from weather pattern prediction and forest fire monitoring to global vegetation analysis. These gridded data sets are distributed mostly as files in HDF, GRIB, or netCDF format and quantify variables like precipitation, soil moisture, or sea surface temperature, along one or more dimensions like time and depth. Although the data cube is a well-studied model for storing and analyzing multi-dimensional data, the GIS community remains in need of a solution that simplifies interactions with the data, and elegantly fits with existing database schemas and dissemination protocols. This paper presents an information model that enables Geographic Information Systems (GIS) to efficiently catalog very large heterogeneous collections of geospatially-referenced multi-dimensional rasters—towards providing unified access to the resulting multivariate hypercubes. We show how the implementation of the model encapsulates format-specific variations and provides unified access to data along any dimension. We discuss how this framework lends itself to familiar GIS concepts like image mosaics, vector field visualization, layer animation, distributed data access via web services, and scientific computing. Global data sources like MODIS from USGS and HYCOM from NOAA illustrate how one would employ this framework for cataloging, querying, and intuitively visualizing such hypercubes. ArcGIS—an established platform for processing, analyzing, and visualizing geospatial data—serves to demonstrate how this integration brings the full power of GIS to the scientific community.