What I did on my vacation: spatio-temporal log analysis with interactive graphics and morphometric surface derivative

Developments in computing and mobile communications mean that massive amounts of data are being produced. MacEachren & Kraak (2001) report that up to 85% of the million terabytes or so of digital data generated every year (Lyman et al., 2000) has a spatial component. This has at least two important consequences for GI science : more spatial data is being generated and new techniques are required to analyse and summarise it ; advantage may be gained from managing data that are not traditionally considered to be spatial in a geographic framework. The research presented here focuses on the former issue. It may result in insight that is of benefit to those undertaking the latter with a view to providing location-based services (LBS) and the like. 1. The Problem The developing influence of the global positioning system (GPS) and various forms of triangulation and other georeferencing associated with mobile telecommunications mean that we are able to collect and are often inadvertently collecting masses of information about our locations in time and space. Plenty of applications exist for these data including : various levels of sophistication of location and user-dependent information provision (where am I? where am I going? what do I know? how familiar am I? what has changed since my last trip?), data synthesis and potentially explanation and prediction. A range of potential measurements exist to help us achieve these aims including : sinuosity, bearing, speed and rates of change. But in order to progress with the new forms of data and derive meaning from them we need to gain an understanding of what we are actually able to collect and how the data might be utilised. We also need techniques for summarising the masses of data that we collect. Here a series of methods for extracting information from the kinds of logs of spatio-temporal activity recorded by mobile navigation and communication devices are presented and used to distinguish clear trends. These include interactive data sifting with the Location Trends Extractor (LTE), the use of feature networks derived from continuous density surfaces of recorded spatio-temporal activity and geocentric parallel plots that relate sequences of spatial locations to multivariate attributes of that space. A number of sample logs have been recorded using GPS receivers. These range in spatio-temporal scope from distinct trips that last for only a few minutes, to a single week’s activity in a space of just a few kilometres radius to a period of a year in which intercontinental travel has occurred. Each contains a sequence of locations with an associated time-stamp. A number of the logs describe periods of vacation! We aim to describe and make inferences about ‘what we did on our vacations’. 2. Visualization : Working Towards a Solution Visualization is a useful tool for providing insight into geographic distributions and affecting ideation MacEachren & Kraak (2001). It is particularly useful at the ‘what have we got?’ stage of an investigation where ideation is important. The LTE software provides a highly interactive graphical interface to logs of the spatio-temporal activity that we have recorded. It has been designed to enable users to gain insight into the such data sets. It can be used to select sections of a log using temporal limits and view the associated space, and importantly vice versa. Linked windows and a range of techniques for interactive querying and symbol shading provide a mechanism for identifying and detecting episodes of activity and repeated events. LTE allows us to split and splice space and time in order to sift through our spatio-temporal records and derive some meaning from them. Indices such as absolute speed, direction, sinuosity and measurements of their variation can be derived using LTE. Graphical selection techniques have been incorporated to identify sections of activity with a particular speed, direction or sinuosity. Sudden changes in spatial co-ordinates, time or any of these derived indices can be used to identify break points between episodes of homogenous spatio-temporal behaviour. Conjectured activity can be measured from these episodes to summarise the likely means by which a user is travelling through time and space. Figure 1 demonstrates.

[1]  Roland Sauerbrey,et al.  Biography , 1992, Ann. Pure Appl. Log..

[2]  Joseph. Wood,et al.  The geomorphological characterisation of Digital Elevation Models , 1996 .

[3]  Valerio Pascucci,et al.  Visualization of scalar topology for structural enhancement , 1998, Proceedings Visualization '98 (Cat. No.98CB36276).

[4]  Stephan Imfeld,et al.  Time, points and space - towards a better analysis of wildlife data in GIS , 2000 .

[5]  A. MacEachren,et al.  Research Challenges in Geovisualization , 2001, KN - Journal of Cartography and Geographic Information.