Use case: Linked Data based thematic web mapping
The world wide web of data makes it possible to have a new kind of web GIS or web mapping: applications that have no prior knowledge of the data they are going to use. It should be possible to visualize and analyse geographical data in the follow-your-nose navigational style that Linked Data makes possible.
Consider a web page with a map window on it. The map window is equipped with common map controls, allowing a user to pan and zoom, to display data on selected objects and to add or remove map layers. By default, the map window shows a general topographic reference map, of which the source is a map tile server. The topographic reference map is composed of images (i.e. raster data) and has a fixed Coordinate Reference System (CRS).
For the map windows no data sets are preconfigured, but map layers can be added from the web of Linked Data. Index datasets are used to find datasets about the theme that the user is interested in. For datasets matching the theme, geographical data are requested, filtered by the current settings of the map window: CRS, zoom level and extent. The requested data are rendered on the map and become objects that the user can select.
More schematically, this is what should be possible in this use case:
- Search for geographical data by theme
- Filter datasets on compatibility with the map
- Request data and draw them on the map
Currently missing in specifications
Specification of geographical extent of a data set
Datasets that are published according to the rules of good behaviour on the web of data will have some metadata. These metadata can give a prospective user an idea of what the dataset is about. Also, the metadata are a good candidate for inclusion in datasets describing datasets and indexation in search engines. A standard for metadata is the VoID.
For this use case it is necessary for a description of the geographical extent of the dataset to be part of its metadata. Common practice in Geo-IT is to specify a rectangle in a certain CRS: the spatial extent. Currently available vocabularies like GeoSPARQL are able to express such a rectangle. What is missing is a standardized way of adding the spatial extent to the (VoID) metadata. Perhaps and extension to the VoID can be suggested?
Specification of the level of detail of a data set
Geographical objects like linestrings and polygons can be described with different levels of detail (generalization levels). The borders of a country, for example, could be described by 10, 100, 1000 or more coordinates. More coordinates mean a higher level of detail. A high level of detail is not always desirable, because it means that datasets become more bulky. In this use case, the appropriate level of detail depends on the zoom level of the map window. To request more data than is possible to render given the size of the map window and the resolution of the display device is a waste of resources and will have a negative effect on application performance.
A specification of the level of detail on in the metadata of a dataset could
- allow data consumers to use level of detail in a filter to select appropriate datasets
- make it easier for data publishers to publish multiple versions of the same dataset with a different level of detail.
The OGC has done some work on a specification of level of detail of a spatial dataset, making it possible to express the level of detail as a single numerical value (reference needed here). Perhaps such a standardized expression of level of detail could be included in the VoID?
Specification of coordinate reference system
Before geographical data can be drawn on a map it is necessary to know the reference system that is used for the coordinate values. The GeoSPARQL specification makes it possible to identify a CRS for data published as Linked Data, but unfortunately the URI identifying the CRS is part of the literal describing the geometry. It is not a separate entity and therefore it is hard to use the CRS as a filter. Filtering on CRS is desirable both in SPARQL queries and in searching in metadata. Data publishers could make muliple versions of datasets or data objects available, with different CRSs. Therefore, being able to specify the CRS is needed for both the object data and the metadata of the dataset.
Perhaps such a standardized expression of CRS could be included in the VoID?