Time has, conceptually, long been considered a symmetrical counterpart of spatial dimensions, although in practice multiple dimensions frequently were relegated to secondary ‘attribute’ status, with temporal characteristics (as well as z as a third dimension) handled as attributes of planar features. Only recently has a full integration into spatial / spatiotemporal data models been pursued.

This is considered a key requirement for dynamic modelling of geospatial processes, with terms like ‘movement’, ‘change’, ‘transfer’ etc. inherently being tied to integrated consideration of time and space dimensions. Early concept applications like the ‘Detroit movie’ (Tobler 1970) followed much later by ‘Flowmapper’ already demonstrated a focus on visualisation and dynamic cartography, leading to a visualisation-oriented development and a lack of sound foundations in data management and process modelling – a clear case of the (visualisation) user interface running ahead of the actual substance of spatiotemporal modelling. In other words, impressive dynamic visualisation was glossing over the fact that underlying structures were only dedicated to and optimised for visualisation, but not analysis and modelling.

Of course visualisation is a key instrument for generating hypotheses, and can very well lead to conceptual innovation. An important starting point was presenting the three (meta-) dimensions of space, time and attribute as equally relevant connected domains, with 2D sections through this cube defining typesof geographical analyses as defined by Brian Berry’s ‘data matrices’ in the early stages of the ‘quantitative revolution’.

Within the framework of these dimensions somewhat different view on similar geometric metaphors of the space-time cube and space-time prism was used by TorstenHägerstrand when first introducing the conceptual and empirical foundations for movementsthrough space as traces and paths in time and space, leading to the concept of time geography.

This approach essentially being an early example of individual-based modelling in social environments, physical processes in space were approached from a general systems theory perspective leading to simulation (Chorley & Kennedy, 1971). Their mathematically and physi­cally founded view on process modelling was implemented only to a limited degree, with work on ‘PCRaster’ being rather the exception than the rule. While Langran (1992) summarizes conceptual views on space-time, ade­quate implementations in current software tools are still rare due to limitations in data models.

Dynamic models of the environment heavily depend on calibration and parameterisation through empirical measurements, which are expensive to conduct over longer periods of time at multiple locations throughout a study area.This has led to a bottleneck in data processing, with traditional databases not able to keep up with the enormous streams of spatiotemporal data. The concept of ‘moving objects databases’ has become particularly relevant in the domains of society and transportation, where increasingly humans and vehicles are serving as sensors, generating enormous quantities of spatiotemporal data and providing a foundation for modelling the flows and dynamics of people, information, goods and assets through space.

Current developments are characterised by two complementary approaches: ‘bottom up’ (‘individual based’, e.g. ‘agents’) simulation of elementary entities and their behaviour aggregating towards groups, regions and ‘societies’, and the ‘top down’ perspective as made popular through the Club of Rome’s world models (Meadows 1972). This system dynamics approach of stocks, flows and feedback structures is designed to help with understanding complex systems over time, and has received only limited attention (and integration) from a spatial sciences perspective.

From an applied and strategic point of view, geospatial information management is increasingly driven by the development of Spatial Data Infrastructures (SDI) as distributed, highly interoperable and partially open services-oriented architectures. These are built upon specifications developed within the Open Geospatial Consortium. These specifications will likely be highly relevant for the practice of geospatial information management in the immediate and foreseeable future, and of course cover temporal characteristics. It is not clear, though, if there is sufficient support for dynamic process modelling beyond rather static views of valid time, time periods, time series, temporal events and other concepts relevant particularly in legal domains.

While time and space, and the modelling of spatiotemporal dynamics have been consistently recognized as an important element of the GIScienceresearch agenda, progress in bridging the gap between con-ceptual understanding and practical implementations and applications was limited, mainly to visualisation-centric solutions.

Particular deficiencies are detected at the interface of systems dynamics and explicitly geospatial models, and with the general design of methods for integrated spatial-temporal analysis. Overarching research questions for PhD topics addressed in this cluster include:

• Which conceptual (and logical) data models and process models are appropriate for time-dynamic geospatial data modelling?
• What are consolidated theoretical and methodological foundations for an efficient integration of sensor data streams into geospatial processing chains?
• How can field- and feature-oriented algorithms be utilized for modelling spatial dynamics?
• How can temporal dynamics be analysed and modelled based on Spatial Data Infrastructure environments?
• How can well-established theories and methods (e.g. concepts of Time Geography) be implemented and developed further for the aggregating population dynamics in space-time based on individual mobility and vice versa?
• How can human factors - including human-computer interfaces, privacy issues, and social and institutional issues – be appropriately addressed in GIScience time and process models?