Geographic information systems (GIS) have been used successfully for a long time to extract insights from static, geo-coded “vector” information such as points, lines, and polygons. However, traditional GIS is hitting scalability limits as a result of the emergence of “mega” Big Data in the form of geo-coded imagery from drones and satellites and time-series Internet of Things (IoT) data. We expect to see a major shift in 2019 to address this issue, driven by the confluence of massive geospatial-temporal data together with advances in Cloud computing, advanced machine learning, and artificial intelligence (AI). New innovations are helping businesses in different industries as diverse as energy and utilities, finance, insurance, and governments capture the full value of this ever-growing, ubiquitous, and vitally-important class of information.
Seeing the larger picture
Today, geospatial-temporal data remains relatively “dark” compared to other kinds of data, in large part due to its size. Consider the fact that some data sets now grow by tens or even hundreds of terabytes per day, making traditional approaches of downloading information and working with geo-coded imagery on a “file-based” level, a daunting quest for obtaining timely insights for geospatial-temporal use cases.
Beyond its complexity and the enormous size of this data, the level of “indexing” for geo-coded imagery from drones and satellites is insufficient. For example, the data from the World Wide Web has been indexed such that we can search or query billions of Web pages within a fraction of a second using a few search criteria or keywords. In contrast, until recently it has not been possible to query petabytes of satellite information rapidly given a few search criteria.
To address these issues, over the past several years a team at IBM Research invented a new technology called IBM Physical Analytics Integrated Data Repository & Serivces, known as IBM PAIRS Geoscope. PAIRS is a Cloud service that provides a ready-made catalog of carefully indexed, diverse, and continuously-updated geospatial-temporal information, enabling scalable access to complex queries and machine learning-based analytics and AI without the need for downloading data.
PAIRS provides four Cloud-based core services. Data or information services provides access to petabytes of up-to-date, AI-ready curated data, such as historical precipitation data and passenger traffic information for all airports in southeast Asia. Data curation services enables clients or content providers to integrate their own data into the PAIRS platform thereby allowing them to exploit, analyze, or monetize their data along with the petabytes of already-curated data. Search or query services enable rapid search given a few specific criteria (e.g., show me all places in the United States where next week’s average forecasted temperature is lower than 40F and the population density is larger than 4,000 people per square mile). Finally, PAIRS provides complete analytics platform services that enable clients to run their own custom analytics without downloading the raw information.
Examples of very challenging questions that can now be answered with PAIRS include everything from “how can we predict the corn yields in a specific region this harvest?” to “where are the most vulnerable places or biggest problems during an earthquake in a specific region?” to “where is the best place to build a wind or solar farm?” to “how can we track the spread of a wildfire?” These queries require advanced analytics, and PAIRS leverages machine learning and AI techniques to make predictions based on a complex mix of parameters, models, and historical data.
2019 will be a defining year for geospatial-temporal data as we finally see the tools needed to exploit this frontier of Big Data come to market.