Visualising Spatial Uncertainty of Social Media Documents

Uncertainty Visualisation is one the current great challenges in cartographic research. The awareness of Cartographers to communicate various kinds of uncertainty has become considerably higher with the emergence of big data. There is no all-embracing methodology to display the nature of uncertainty, as a useful visualisation depends strongly on the type of uncertainty, the spatial data feature type as well as the source of uncertainty. This work tackles two location uncertainty problems within case studies that commonly appear when mapping geo-tagged microblogs of social media; (1) the communication of multiple uncertainty levels of georeferenced microblogs; and (2) the indication of spatial uncertainty caused by generalising distributions. While the latter case study empowers the user to intuitively depict the centres of gravity and the major orientation of multiple geo-tagged text document sets, both visualisation proposals allow a coincident visualisation that indicate positional uncertainty and semantic information such as keywords on one map face, which is particularly suitable for laymen. In the past years, many related works address to analyse real world events and to reveal spatial patterns from social media such as Twitter. However, users publish tweets in different locational accuracy levels. Geo-tagged tweets are associated either directly with coordinates by a GPS enabled device, or, in a more inaccurate level, with a place (e.g. city) that is defined by a regional bounding box polygon. When geo-tagged tweets of contrasting accuracy levels are mixed in processing and display the need for indicating the uncertainty level becomes necessary. In this work, symbolisation strategies are to be presented for single depiction and aggregated sets of tweets that enable the user to understand the contrasting locational accuracy levels. Furthermore, the uncertainty following generalisation processes is described. Large quantities of geo-tagged text documents have to be generalised within the visualisation process in order to keep a visualisation understandable and free from clutter. Commonly used spatial temporal clustering algorithms are extended by a semantic dimension that express the text similarity between tweets. The results are presented in form of the error ellipse, a well-known statistical graphic tool, which visually encodes statistics as a compact distributional summary. Each cluster is represented by one error ellipse. The form of the drawn ellipse shows the cluster´s centre of gravity, the major orientation of the distribution as well as indicates the extent of the geo-tagged text documents occurrence. The ellipse´s surface is graphically encoded with high incidence text keywords. These strategies of displaying the positional uncertainty of social media documents are presented by analysing Twitter microblogs during the Munich Oktoberfest.     «

Uncertainty Visualisation is one the current great challenges in cartographic research. The awareness of Cartographers to communicate various kinds of uncertainty has become considerably higher with the emergence of big data. There is no all-embracing methodology to display the nature of uncertainty, as a useful visualisation depends strongly on the type of uncertainty, the spatial data feature type as well as the source of uncertainty. This work tackles two location uncertainty problems within...     »