Spatio-temporal Analytics on IBM Physical Analytics Integrated Data Repository and Services (PAIRS)

View publication


IBM Physical Analytics Integrated Data Repository and Services (PAIRS) is a platform, specifically designed for massive geospatial-temporal data (maps, satellite, weather, drone, IoT), query and analytics services. Geospatial-temporal data acquisition and preparation can be challenging as there exist vast variety of data formats and geo-projection methods for geospatial-temporal data such as HDF, GRIB, netCDF or GeoTiff, etc. These binary file formats are optimized for data storage efficiency, however, it is highly inefficient for spatial-temporal analysis, where data across spatial locations and temporal ranges are required to be processed simultaneously where parsing of hundreds or thousands of such binary files is computation intensive, none the less posing heavy burden in data analytics research and system implementations. To address the afore-mentioned challenges, PAIRS has been developed to provide search-friendly ready access to a rich, diverse, and growing catalog of historical and continuously updated geospatial-temporal information, on a collection of weather forecast models, geoscientific model outputs, satellite imagery, aerial survey data, measurement data collected through IoT (internet-of-things), etc. PAIRS is a multi-peta bytes data platform and ingesting multi-tera bytes of data on a daily basis. PAIRS provides friendly query from across multiple datasets, where the dataset may come with different spatial and temporal scales, for example: “Show me all urban areas where it will be sunny for the next 10 days and where the population density is larger than 500 people per square mile and where there are at least two coffee shops per one square mile area”. PAIRS also support calculation on the source data on the fly, for example: calculating the radioactive heat loss from Landsat 8 satellite data. In this talk, we will show a few examples of remote sensing analytics, which were developed in short periods of time with relative ease by leveraging the PAIRS data capability. PAIRS is currently being used as key data analytics platforms at our clients and partners, and at the meanwhile PAIRS team is also utilizing PAIRS to develop spatiotemporal analytics capabilities such as solar forecast, long term weather forecast, change detection, vegetation management and etc. In this talk, we will talk in some detail the spatiotemporal analytic models developed for these applications.