The Core Concepts of Spatial Information are designed to facilitate spatial computing and reduce its complexity. They also serve as conceptual lenses on environments, allowing for different perspectives on them, fed by data with any sort of spatial reference. We specify the Core Concepts as Abstract Data Types (ADT), defining a set of core computations for each concept, through which users can ask spatial questions (Kuhn & Ballatore, 2015). The ultimate goal is a generic Application Programming Interface (API) for spatial computing.
The main motivation of this project is to promote transdisciplinary research through a more intuitive access to spatial data and computing (Kuhn, 2012). Spatial computing is seen as an enabler, but remains notoriously complex, especially for those without expertise in GIS. A large part of this complexity results from historically grown command sets rather than from inherent difficulties. The Core Concepts and Computations constitute a high-level language that allows for question-based spatial computing across disciplines. Spatial computations get organized around questions, instead of being accessed through procedural commands, limited to certain file formats, and requiring frequent format conversions.
A case study (Vahedi et. al, 2016) exemplifies the gain from applying the Core Concepts to spatial analysis. An economist studying economic activity in China decides to use nighttime light as an indicator. His goal is to quantify nighttime light within a 50-kilometer buffer around Chinese road networks and excluding gas flares. He develops a lengthy ArcPy script solving the problem in around 10 steps (see http://economics.mit.edu/files/8945). To answer the same question through the Core Concepts, one conceptualizes nighttime luminosity as a field. In a single computational step, one restricts the field domain to a 50-Kilometer buffer around roads excluding gas flares, and coarsens the granularity of the field.
The Core Concepts get developed and tested through a four-layer architecture. We assume (and willhave to test) that domain specialists from any discipline can usefully express their spatial questions (Domain questions layer) in terms of Core Concepts (Core Concepts layer). A Mediation layer then automatically translates the results into commands of existing spatial technologies (Technological layer). Together, the Core Concepts and Mediation layers act as a wrapper around existing spatial computing technologies, such as GIS or statistical packages with spatial data structures and analysis functions.
As of now, the set of Core Concepts consists of seven concepts, namely,
– One Base Concept: Location
– Four Content Concepts: Field, Object, Network, Event
– Two Quality Concepts: Granularity, Accuracy
It appears likely that the Base and Content Concepts are sufficiently complete to cover the scope of most GIS analyses. The number of Quality Concepts, on the other hand, is likely to grow to include such ideas as the provenance of spatial data.
The Core Concepts and their operations are specified in Haskell. The mediation layer is currently being implemented using Python and GDAL, generating wrappers for translating spatial questions into existing spatial technologies. The latest status of implementation is always accessible as a generic API programming library with Haskell and Python implementations.
Validation and Future Work
We validate the Core Concepts, testing their domain neutrality and suitability by systematically redoing GIS analyses (such as bird flu risk assessment, solar panel placement, industrial activity monitoring etc.).
The challenges we are addressing include:
- Determining the necessity of other concepts (e.g., provenance);
- Completing the set of operations associated with each concept;
- Using core concept lenses to design data products;
- Re-designing the teaching of GIS around the Core Concepts;
Challenges to address in the future include:
- Understanding spatial intuition and assessing the concepts’ effectiveness in aiding this intuition;
- Assessing the potential impact on end user interfaces of GIS;
- Implementing the Core Concept API in the form of cloud services;
- Designing RDF vocabularies to represent core concept instances as linked data;
This work started with the observation that Geographic Information Science lacks a set of core concepts (cf. the ideas of a cell in biology or of value in economics). A sequence of early publications (Kuhn 2011, Kuhn 2012) proposed slightly different lists of Core Concepts, with subsequent research leading to the current list. The reason to eliminate the concept of neighborhood (contained in these publications) was that it cuts across the other content concepts and is part of the idea of location. The reason to eliminate meaning and value is that there are no solid theories for them yet to make them
“core” to spatial analysis.
Contact: Werner Kuhn (Center for Spatial Studies)
Researchers: Thomas Hervey, Sara Lafia, Behzad Vahedi
- Publications (including links to slides):
- Allen, C., Hervey, T., Lafia, S., Phillips, D., Vahedi, B., Kuhn, W. (2016). Exploring the Notion of Spatial Data Lenses. Ninth International Conference on Geographic Information Science. (accepted)
- Kuhn, W. (2012). Core concepts of spatial information for transdisciplinary research. International Journal of Geographical Information Science, 26(12), 2267-2276.
- Kuhn, W. (2011). Core Concepts of Spatial Information: A First Selection. XII GEOINFO, November 27-29, 2011, Campos do Jordão, Brazil (pp. 13–26).
- Kuhn, W. & Ballatore, A. (2015). Designing a Language for Spatial Computing. Lecture Notes in Geoinformation and Cartography 2015, AGILE, Lisbon, Portugal, pp 309-326. Best Paper Award. PDF
- Vahedi, B., Kuhn, W., Ballatore A. (2016). Question-Based Spatial Computing – A Case Study. Lecture Notes in Geoinformation and Cartography (AGILE 2016) (pp. 37 – 50). Berlin: Springer. PDF