SOLAP Model For Multidimensional Aggregation And Analysis Of Remotely Sensed Information

From the70s of last century, a series of consecutive projects on Earth observation have generated a large amount of historical remotely sensed dataset in time series. Recently, more and more remote-sensing satellites were launched and the technologies of high-resolution (spatial\temporal\spectral) Earth observation were improved considerably. The amount of remotely sensed data were explosively growing with the characteristics of "big data"(high Volume, Variety and Velocity), and became the key data source of the GIS (Geographic Information System). On the other hand, the spatial OLAP (On-Line Analytical Processing) introduces the OLAP technique in Business Intelligence (BI) to spatial data based on the spatial data warehouse. SOLAP provides a powerful tool for multi-view, multi-theme, multi-granularity and multi-level information exploring via the cube model for spatio-temporal data. However, current researches and applications are still focus on the vector spatial data, therefore, how to integrate the remotely sensed data that characterized by the "big data" into the SOLAP cube is an opportunity as well as challenge for SOLAP. To address the issues in the existing researches on the SOLAP for field data, this paper aims at the high-performance SOLAP cube for remotely sensed data, and carries out the following key researches:(1) Based on the concepts, approaches and architectures of SOLAP, present the SOLAP cube model "TileCube" for remotely sensed data based on the geographic grid. First, the spatial measurements based on multi-level geographic grid are extended into the traditional SOLAP cube. The proposed model built in a distributed environment presents the "multi-dimension" characteristic of remotely sensed data reasonably, which includes the supports in multi-spatial hierarchies and several non-spatial dimensions, and the definition of the aggregation relations among different cubes. Moreover, TileCube introduces the Multidimensional Map Algebra (MMA) into the expression of the aggregations between levels in hierarchies and aggregations among the cubes. Finally, the scalable storage mechanism and query method of TileCube are studied based on BigTable from physical layer, and the mapping method is implemented from physical layer to logical layer in the model.(2) For the performance issue of large-scale remotely sensed data processing in the TileCube, this paper proposes a Map-Reduce approach for high-performance aggregations involved in the cube building and analyzing. First, to accelerate the ETL of data cube, the gridding method of the time-series remotely sensed data stream is presented as well as its optimization. Then, the Map-Reduce-enabled MMA is proposed to decompose the large-scale aggregations into multiple parallel map algebra tasks based on the tile, and the SOLAP operations, such as Roll-Up and Drill-Across, are implemented via the MMA. Finally, the work mechanism of SOLAP engine is illustrated which provides a powerful driver for SOLAP analysis.(3) For the purpose of verifying the proposed approaches, the drought monitoring and analyzing prototype system is built from the SOLAP perspective, and the long time-series data from MODIS (Moderate Resolution Imaging Spectroradiometer), the real-time data from ground station and the basic spatial data are integrated into the TileCube. In the prototype, the spatio-temporal aggregations for the drought information are implemented, and the ability of interactive and multi-dimensional analysis is presented via several drought queries examples.The original contributions of this paper are as follows:(1) Propose a distributive SOLAP cube model which resolves the "multi-dimension" description issue of remotely sensed data in the model.(2) Present a parallel multidimensional map algebra via Map-Reduce computing paradigm that enables the spatio-temporal aggregations and online analysis for large-scale data.(3) Investigate a cube gridding method of time-series remotely sensed data stream, accelerating the ETL process of SOLAP cube.In line with the technical roadmap of "basic theory—cube model computation/analysis—application/testing", this paper has researched on the theories, methods and implementations of the SOLAP for remotely sensed data, addressed the issues of "multi-dimension" and "big data" in the TileCube. These achievements not only help to promote the development and application of SOLAP in the field of GIS, but also provide strong technical support for the further researches on the high-performance Geospatial Cyber-infrastructure.