Geospatial semantics, ontology and knowledge graphs for big Earth data

Big Data has attracted a lot of attention from governments, industries and academia, and it has been applied to a large number of fields around the world. Big Earth data refers to big data associated with the Earth sciences that is characterized as being massive, multi-source, heterogeneous, multi-temporal, multi-scalar, highly dimensional, highly complex, nonstationary, and unstructured (Guo, 2017b; Guo et al., 2017a). Most big Earth data is related to a geographical location and is usually referred to as geospatial data (Lee & Kang, 2015). Geospatial data is not only an important component of and the main way of organizing and visualizing big Earth data, it is also the foundation for integrating multisource, heterogeneous big Earth data. With the development of Earth observation, deep exploration, computer simulations and other technologies, the capacity to acquire geospatial data has grown very fast, and thus the volume of geospatial data is also increasing exponentially. For example, according to the Union of Concerned Scientists Satellite Database (UCS (Union of Concerned Scientists), 2019), there are 2062 operational satellites currently in orbit around the Earth. These satellites produce huge amounts of remote sensing images with various spatial, temporal and spectral resolutions. Since the 1950s, in order to promote the full sharing and value-added utility of geospatial data, developed countries as well as some developing countries have launched a number of geospatial data-sharing initiatives and programs that have established many spatial data infrastructures (SDI), data centers, and data-sharing platforms around the world (Bai & Di, 2010; Goodchild, Fu, & Rich, 2007; Harvey & Tulloch, 2006; Hu, Janowicz, Prasad, & Gao, 2015) and great achievements have been made in geospatial data sharing. For instance, the Data Sharing Service System of CAS Earth (http://data.casearth.cn/) launched by the Chinese Academy of Sciences has integrated 5.02PB data resources since 2018, and the National Integrated Earth Observation Data Sharing Platform of China (http://www.chinageoss.org) has collected 2,140,000 scenes acquired by 9 land-observation satellites, 1,440,000 scenes from meteorological satellites, and 10,000 scenes from ocean-observation satellites. However, for most geospatial data-sharing projects, top-down organizational mechanisms are usually adopted, which results in networks with large amounts of data remaining in the hands of individual scientists and not being effectively shared. In terms of technology, the existing geospatial data-sharing platforms mainly use simple keyword-matching to search metadata, which leads to incomplete and inaccurate search results due to the lack of semantic reasoning. Therefore, in terms of the mechanisms used, geospatial data-sharing is moving towards a combination of top-down and bottom-up mechanisms and, in technical terms, towards precise data searches and BIG EARTH DATA 2019, VOL. 3, NO. 3, 187–190 https://doi.org/10.1080/20964471.2019.1652003