Dynamic and Mobile GIS: Investigating Changes in Space and Time - Chapter 4

Part II Modelling Approaches and Data Models Part II of the book ‘Dynamic and Mobile GIS’ focuses on modelling approaches, especially those appropriate to depicting dynamic processes in GIS. The four contributions present a high level of novelty and suggest a dramatic evolution of data models, spatial analyses and spatial queries. As opposed to initial GIS approaches considering geographic features and relations among them as independent of time, Kate Beard in Chapter 4 proposes an event-based approach in which change itself is the central concept that is modelled. Events as explicit representations of change with associated attributes of change such as rate of change or rate consistency provide the key units for exploration and analysis of the mechanism of change. In this approach, the time dimension dominates the spatial dimensions as the ordering of events in time is critical. An event-based view provides the foundation for the analysis of dynamic phenomena and is therefore naturally appropriate for dynamic GIS. Chapter 4 includes a very complete state-of-the-art presentation of dynamic process modelling and event-based models. A formal categorization of changes is then proposed, sources of events are described and a method for event visualisation and exploration is outlined. Femke Reitsma and Jochen Albrecht present in Chapter 5 a new process-based data model called nen (after node-edge-node graph representation). While most of the existing theories and models for simulating processes focus on representing the state of the represented system at a moment of time, this approach expresses and represents information about processes themselves. This data model provides advantages in querying, analysis and exploration of process descriptions under computer simulation conditions – or in silico. Chapter 5 includes a description of the model and presents its application in a small watershed modelling test case. Part of the originality of nen is to provide a new epistemological window on the modelled results, allowing for new process-oriented queries and analyses. It allows questions to be asked (where is a process operating at a particular instant of time? how has the process changed over time? etc.) not directly answerable with current object-centred formulations that focus on the states of a system resulting from the operation of a process. In Chapter 6, Muki Haklay extends the comparison between Map Calculus and Map Algebra in the context of dynamic raster GIS. Map Calculus is an alternative to current representation in GIS, and is based on the use of function-based layers in GIS. Its main strength is its ability to treat analytical layers in their symbolic form in a similar way to the manipulation of mathematical functions in software packages. This chapter focuses on the particular challenges of dynamic modelling in GIS, exploring the ways in which it is implemented in Map Algebra and outlining © 2007 by Taylor & Francis Group, LLC 54 Dynamic and Mobile GIS: Investigating Changes in Space and Time how such models can be implemented in a Map Calculus-based system. It appears that Map Calculus allows easier linkage to rapidly changing inputs and easier implementation of dynamic models based on differential equations. The use of Map Calculus poses certain challenges such as the reformulation of common GIS operators and the consideration of optimal visualisation methods. Ultimately, it should make GIS more accessible to domain experts, as they can focus on the construction of the model and not on finding ways to fit a conceptual model within the constraints of GIS. Finally, Peter van Oosterom explores, in Chapter 7, issues related to spatial constraints in data models. In GIS, constraints are conditions that must always be valid for the model of interest. In a dynamic context, with constantly changing geo-information, any changes arising should adhere to specified constraints; otherwise inconsistencies (data quality errors) will occur. Constraints should be part of the object class definition, just as with other aspects of that definition, such as attributes, methods and relationships. Currently, the implementation of constraints (whether at the front-end, database level or communication level) cannot be driven completely automatically by constraints’ specifications within the model. This chapter demonstrates the need for the integral support of constraints through four quite different cases but all dealing with dynamic situations: a VR system for landscape design, cadastral data maintenance, topographic data maintenance and a Web feature service. It proposes a complete description and classification of constraints and describes some solutions for the formalisation and the implementation of constraints in the different presented cases. © 2007 by Taylor & Francis Group, LLC Chapter 4 Modelling Change in Space and Time: An Event-Based Approach Kate Beard Department of Spatial Information Science and Engineering, University of Maine, Orono, USA 4.1 Introduction Initially geographic information systems modelled geographic features and the relations among them under the assumption that such features were independent of time. A rationale for this perspective is that many geographic features retain their identity and location for long periods of time. Given the persistence of these fundamental properties, representation of change was not an initial consideration for geographic information systems. Additionally, early spatial data collection methods (primarily photogrammetry) focused on capturing these fundamental properties: identity and location, but were generally too expensive to repeat with a frequency that could support interesting change analysis. More recent GIS research has begun to address models for representing the dynamic and mobile components of geographic features. What has remained constant, however, is that geographic layers or features are the central units of analysis and GIS operate on these units. In an event-based approach geographic features or locations are not the primary focus. This chapter describes an event-based approach in which change itself is the central concept that is modelled and change units are the principal objects and units of analysis. In this event-based approach the time dimension dominates the spatial dimensions as the ordering of events in time is critical. Because the principal unit of analysis and the organising dimension are fundamentally different, new approaches are needed for change objects. Section 4.2 of this chapter discusses previous event-based models and approaches that have appeared in the literature. Section 4.3 presents a categorisation of change, Section 4.4 describes the proposed event model, Section 4.5 describes sources of events, Section 4.6 outlines a method for event visualisation and exploration and Section 4.7 concludes the chapter with a summary and future research challenges. 4.2 Previous approaches to time and dynamic models in GIS Beginning in the late 1980s and early 90s GIS research began to address the dynamic aspects of geographic features and to include time in GIS (Armstrong, ____________________________________________________________________________________ Dynamic and Mobile GIS: Investigating Changes in Space and Time. Edited by Jane Drummond, Roland Billen, Elsa João and David Forrest. © 2006 Taylor & Francis © 2007 by Taylor & Francis Group, LLC 56 Dynamic and Mobile GIS: Investigating Changes in Space and Time 1988; Langran, 1992). The addition of time-supported tracking the history of spatial objects and their attributes and the prediction of potential future behaviour and change. The primary focus however remained the geographic feature with the temporal dimension added in the form of a time stamp to monitor and analyse successive states of these features (Abraham and Roddick, 1999). In much of this work the central unit of analysis was viewed as the spatial layer or theme and change was conceived of as modifying the fabric of a layer (Langran, 1992). This view is generally associated with the concept of the spatial snapshot. The influence of the object-oriented view shifted the association of change to individual objects. Worboys (2005) terms this the object change view. Under this view an object has a unique identifier that it maintains while changes may occur to its spatial and non-spatial properties. A related pre-GIS view was introduced by Hagerstrand (1970) who was interested in analysing the dynamic behaviour of individuals in geographic space. He was particularly interested in the movements of individuals through space and time, a perspective which lead naturally to an object-based view focused on the individual. Hagerstrand’s work has been recently promoted by others (Mark, 1998; Miller, 2003) under the concept of geo-spatial lifelines. Hornsby and Egenhofer (2000) address the object change view generally under the concept of identity-based change. A particular challenge for the object change view lies in establishing and maintaining the identity of an individual object over time. Questions naturally arise as to when and what change becomes so substantial that an object is no longer the same object. Noting limitations in models that simply add time stamps to support the management of versions and state changes to geographic features or locations, several researchers proposed event-based models (Peuquet and Duan, 1995; Claramunt and Thériault, 1995; Worboys and Hornsby, 2004; and Worboys, 2005). Event-based models move from geographic feature identification and location characterisation to an explicit focus on change. Within the event-based view subtly different definitions and perspectives on events exist. Events are variously described as happenings or expressions of change. Claramunt and Thériault (1995) describe events as things that happen. More specifically they note processes lead to changes in entity states and these changes show the result of the process and constitute events. Peuquet (1994) describes an event as denoting some change in some location(s) or some object(s) and Peuquet and Duan (1995) describe an event as representing the spatiotemporal manifestation of some process. These variations in the notion of an event compare with the variability in event definitions generally. The SHOE General Ontology (SHOE) defines an event as something that happens at a given place in time. In the specifications of the Dublin Core metadata standard, an event is defined as ‘a non-persistent, time-based occurrence’. Quine (1985) described events as objects where objects are regions bounded in space and time. He further notes that events can be broken into sub-parts and arranged in a taxonomic hierarchy. The common factor among the event-based views is the requirement that change be explicitly modelled and they share the objective of facilitating the analysis of © 2007 by Taylor & Francis Group, LLC 4. Modelling Change in Space and Time: An Event-Based Approach 57 change, patterns of change, or happenings through time. Each professes the importance of change or happening as the central unit of analysis. The Event-Based Spatiotemporal Data Model (ESTDM) proposed by Peuquet and Duan (1995) temporally orders changes to locations within a pre-specified geographical area. It is interesting to note that while change (an event) is explicit in this model it is subtly subservient to a time location and a spatial location (i.e. bound to pixels). Specific changes are associated with a stored temporal location ti called a time stamp in an event list. Furthermore each event list is associated with a single thematic domain (layer). The ESTDM (Peuquet and Duan, 1995) identifies itself as an event-based data model yet it does not fully support change as the primary unit. The difference is that the focus remains on a value change associated with a location (a pixel) and that change is not the characterised unit. In a pure event view the change unit itself is characterised not the geographic feature or location. Worboys (2005) and Worboys and Hornsby (2004) describe an event as a happening or occurrent to be distinguished from a thing or continuant. They note the weakness of snapshot models as lacking explicit representation of events and explicit representation of change. They argue that events are needed to capture the mechanism of change but describe events more as happenings, activities and processes, rather than explicitly characterised change units. In an object change view we would typically record changes in the properties of an object. Assume that a house is repainted from white to yellow and hence undergoes a non-spatial change to its colour property. The primary object is the house and we would record its new colour, possibly keeping a record of its previous colour. In the event view, ‘painting the house’ is the recorded unit along with its specific properties such as its time of onset, its duration and perhaps the method by which it was accomplished. In an event view the focus moves from the change to a particular house to an analysis of the change objects themselves, e.g. an analysis of painting events through for example a comparison of their durations, seasonal patterns, differences in methodology. The difference in these perspectives can perhaps be illustrated by considering an example query. Using a GIS, a user might ask for all the houses that were painted over the last year and receive a map showing the geographic distribution of these houses. The focus is on a geographic object and a particular type of change to that object type. In an event view the request would be for the ‘house painting events’ over the last year. In this case the set of house painting events would have the same geographic distribution but clearly different sets of attributes that could be queried and analysed. There were early calls for maintaining records of events and processes as the basis for understanding dynamic behaviours (Chrisman, 1998), but the realisation of the event view owes much to new technologies that are now able to deliver a wide range and volume of spatiotemporal information. Environmental monitoring and sensor data streams are creating repositories of information with high temporal resolution that support the analysis of change. Fine temporal resolution data streams begin to provide a picture of how processes operate and a foundation for investigating cause and effect. Because the data are now becoming available to truly © 2007 by Taylor & Francis Group, LLC ... - tailieumienphi.vn