Ontology of core concept data types for answering geo-analytical questions

Abstract

In geographic information systems (GIS), analysts answer questions by designing workflows that transform a certain type of data into a certain type of goal. Semantic data types help constrain the application of computational methods to those that are meaningful for such a goal. This prevents pointless computations and helps analysts design effective workflows. Yet, to date it remains unclear which types would be needed in order to ease geo-analytical tasks. The data types and formats used in GIS still allow for huge amounts of syntactically possible but nonsensical method applications. Core concepts of spatial information and related geo-semantic distinctions have been proposed as abstractions to help analysts formulate analytic questions and to compute appropriate answers over geodata of different formats. In essence, core concepts reflect particular interpretations of data which imply that certain transformations are possible. However, core concepts usually remain implicit when operating on geodata, since a concept can be represented in a variety of forms. A central question therefore is: Which semantic types would be needed to capture this variety and its implications for geospatial analysis? In this article, we propose an ontology design pattern of core concept data types that help answer geo-analytical questions. Based on a scenario to compute a liveability atlas for Amsterdam, we show that diverse kinds of geo-analytical questions can be answered by this pattern in terms of valid, automatically constructible GIS workflows using standard sources.

Loose programming of GIS workflows with geo‐analytical concepts

“…This distinction is seldom drawn, yet we believe it is required to capture how concepts can be represented by different geodata formats. In a nutshell, our idea is that, whenever analysts compose workflows, they interpret data in a way that adds missing semantic information to make effective use of the data (Scheider, Meerlo, et al, 2020). Furthermore, task specifications in the form of explicit application constraints (e.g., “perform an operation of type X”) are not supported, and systematic validations of workflows are still lacking.…”

“…The implementation of this approach in APE was inspired by the PROPHETS loose programming framework (Lamprecht et al., 2010; Naujokat, Lamprecht, & Steffen, 2012) and uses the same underlying semantic linear‐time logic (SLTL) synthesis method (Steffen, Margaria, & Freitag, 1993). The approach has already successfully been used in different bioinformatics (Lamprecht, 2013; Palmblad, Lamprecht, Ison, & Schwämmle, 2018) and geoinformatics (Al‐Areqi, Lamprecht, & Margaria, 2016; Kasalica & Lamprecht, 2019; Scheider, Meerlo, et al, 2020) applications. For example, in Kasalica and Lamprecht (2019) we used APE to synthesize cartographic workflows for the automatic creation of maps depicting bird movement patterns in the Netherlands.…”

“…We used tools from the GMT (Generic Mapping Tools; Wessel, Smith, Scharroo, Luis, & Wobbe, 2013) collection as the basis for the workflow composition. For another example, in Scheider, Meerlo, et al (2020) we used the CCD ontology to compute a livability atlas of Amsterdam using standard GIS operations as provided by ArcGIS. While all these applications were very promising, an empirical study of the workflow synthesis quality has yet to be done.…”

“…In this section we summarize our previous work on the CCD ontology, which provides the semantic basis for this workflow construction study. The lightweight ontology CCD was introduced in Scheider, Meerlo, et al (2020) and is formalized in the Web Ontology Language (OWL) 1 . CCD defines data types as intersections of OWL classes representing combinations of geo‐analytical concepts from four semantic dimensions: (a) geometric layer types , which generalize geometric properties of layers; (b) core concepts of spatial information (Kuhn, 2012), which capture what these layers represent; (c) measurement levels of attributes; as well as the notion of (d) extensiveness .…”

“…A basis for building such a semantic model are core concepts of spatial information , which were suggested by Kuhn (2012) as a generic interface to GIS, functioning as conceptual lenses an analyst may use to interpret spatial features. The core concepts were recently formalized into the core concept data type (CCD) ontology , which is a semantic data type system which captures the different ways in which core concepts can be represented in geodata at different levels of measurement (Scheider, Meerlo, Kasalica, and Lamprecht 2020). CCD types allow us to add necessary geo‐analytical constraints and capture functionality in terms of transformations in four semantic dimensions.…”