3. Broadening the Scope of Geovisualizations
In this section, we discuss how geovisualizations can be constructed with abstract visual representations without the aid of cartographic maps. We have discussed about geovisualization exclusively for spatial data in Section 2. This can be extended to geospatiotemporal data, which is the type of data readily available today. Hence, the visualization of four-dimensional (4D) data is a relevant topic to discuss. Considering the prevalence of geospatial data with multiple attributes, we discuss some of the multivariate visual representations which are used in geovisualization.
3.1 Abstract Visual Representations
The visualizations discussed so far consider the maps in scale, which one can deviate from, when using Cartograms. Cartograms are graphics which map attribute properties to area of entities with boundaries, e.g. counties or districts, states, countries, etc. Though the Cartogram is generated from a cartographic map, it does not necessarily honor geographic precision, as demonstrated by its different types, in Figure 4.
Despite the presence of geo-reference data or metadata, not all geovisualizations are built with the reference of or in the context of a cartographic map, as shown by examples of metro map and volume visualization of geological models in Figure 5.
Abstractions such as schematic diagrams of transport networks and metro maps , serve purposes other than geo-references. In 1933, Harry Beck used “navigation through the London Underground” as a motivation for the design of the London Underground Map, popularly known as the Tube Map, today. Similarly, geological modeling and its visualization do not require the context of cartographic maps for understanding the geological processes. Visualizations of indoor of buildings or man-made habitations, e.g. for movement data in the building, can use abstractions based on floor plans. The floor plans may be considered to be maps, however, for most of the applications, they are used without geo-references. As an example, Lanir et al.  have used tangram diagrams to visualize the spatio-temporal behavior of museum visitors.
3.2 Geospatio-temporal Data Visualizations
While geospatial datasets pertain to two- or three-dimensional space, these datasets predominantly include temporal variations as well, making them geospatio-temporal datasets. Temporal aspects of the data are more effectively captured using animation in comparison to other techniques for visualization of four-dimensional (4D) datasets.
Fig. 5: Geovisualizations without the reference of cartographic maps. (Top) London Underground Map, or the Tube Map, is a schematic diagram which enables navigation through the London Underground system. (Bottom) Geological visualizations involve studying geometries or surface/volume topology involved in geology, such as folds, faults, etc.
(Image courtesy: https://tfl.gov.uk/maps/track/tube and http://app.visiblegeology.com/)
Adrienko et al.  have catalogued different methods used for exploratory analysis of spatio-temporal data. Their study maps different types of datasets and visualization tasks to corresponding types of geovisualizations and user interactions, for effective data analytics. For example, map animation is useful for time stepping for all types of data. However, specifically for existential changes (related to actual events) and location changes, space-time cube is used. A space-time cube is a 3D projection of the 4D data, where one spatial dimension, usually height, is not included. Space-time cubes are useful for tasks which do not require studying height changes; whereas, animation is the method to use for observing height changes, as it uses all of the 4D data.
With regard to map animation, Harrower and Fabrikant  have sensitized the need for application of appropriate design principles for dynamic (i.e. time-varying) visualizations. They have asserted the need for research in not just cartography, but also in animation and human cognitive science to make effective map animations. Map animations can be applied to all visualizations described so far. For example, understanding the mirror neuron system in neuroscience research helps in understanding the effectiveness of dynamic visualizations , as the mirror neuron system plays a key role in observational learning in motor tasks. A good example of map animation is the use of timelapse to demonstrate the changes on the Earth during 1984-2016, using the Google Earth Engine 3.
Bach et al.  have proposed the use of a generalized space-time cube data structure, which is the conceptual representation of time with respect to space. The visualizations are generated using different types of operations on the cube, which are effectively dimensionality reduction processes for converting three-dimensional cubes to readable two-dimensional visual representations. While the cube is applicable to geospatio-temporal data, its use can be extended to other datasets such as videos, networks, etc. The operations include space shifting, time flattening, time juxtaposing, etc., which may be a singleton or multiple operations. The humble origins of the term space-time cube comes from the definition of the term, “time geography” by Torsten Hägerstrand in 1970 , which is the “a time-space concept” for developing a kind of socio-economic web model.
3.3 Multivariate Geovisualization
Most GIS (geographic information science) datasets acquired today contain more than one variables. e.g. for study of land-use, one considers three categories of variables, namely, elevation, edaphic factors, and climatic factors, which altogether are nine variables . The edaphic factors include plant-available water capacity, soil organic matter, etc.; and the climatic factors include mean precipitation, degree-day heat sum, etc. during growing season. Another example of such multivariate geospatio-temporal data is a household survey dataset for assessing public health programmes. Survey data has as many variables as the number of questions in the survey questionnaire .
between the usage of “vertices” and “edges” for the data structure, and “nodes” and “links” for the network data itself.