Object-Oriented Design for Temporal GIS

Monica Wachowicz

• Publisher: CRC Press
• Year: 1999
• ISBN:
  9781000687187
• Edition: 1
• Topic: Informatique
• Subtopic: Programmation orientée objet

There has been an increasing demand in GIS for systems that support historical data: time-series data as well as mobility information. From a modelling perspective, there are advantages in integrating object-oriented analysis and design to databases as well as to visualisation capabilities of GIS.
Object-Oriented Design for Temporal GIS explores

CHAPTER ONE

Introduction

Geographical information science has recently emerged as a distinct interdisciplinary knowledge field involving many diverse areas such as geography, cartography, engineering and computer science. In this field, geographic information systems (GIS) have been used for analysing spatio-temporal data sets pertaining to social, environmental and economic studies. This has led to the integration of a variety of socio-economic and environmental models with GIS. Examples include the innovative GIS-based monitoring model developed by Blom and Löytönen (1993) to monitor current epidemics in Finland, including HIV. This model integrates spatial diffusion, spatial interaction and environmental modelling into a GIS-based model for monitoring the passing of infectious diseases between individuals. The goal of this model is to provide disease-specific forecasts for the future course of an epidemic.

The European Groundwater Project (Thewessen, Van de Velde and Verlouw, 1992) is one example of the integration of existing non-spatial simulation models with spatial data sets. The result is the design of a GIS-based environmental model that provides rapid and coherent access to the most significant causes and effects of groundwater contamination. Physical and chemical models have been integrated into the GIS-based model so it can identify serious threats to the quality and quantity of groundwater resources in the European Union.

The integration of the CLUE model (conversion of land use and its effects) with a GIS is an example of a dynamic, multi-scale, land use change model developed to explore the complexity of the interactions between socio-economic and biophysical factors in land use changes. It was applied to data from China, Ecuador and Costa Rica (Verburg et al., 1997). The results indicate the importance of understanding the dynamics of land use within a multi-scale scenario. Implementation of such a model was essential to explore the spatio-temporal patterns of land use change under different scenarios of population growth and food demand.

Researchers and developers are continually uncovering different uses for GIS-based models in non-traditional applications. Burrough and Frank (1995) draw attention to the diversity of ways of perceiving the same knowledge domain, and consequently the proliferation of many models for handling the knowledge domain at different levels of complexity as well as aggregation in GIS. The study of common concepts and principles among these models is essential when formulating design criteria and strategies to support and advise users on how to integrate them in a GIS. An array of possibilities and new perspectives are expected to arise on how this could be achieved. This book proposes the object-oriented paradigm as a common framework to handle the complexity of semantics of spatio-temporal data defined within a knowledge domain.

1.1OBJECT-ORIENTED ANALYSIS AND DESIGN

Object orientation in modelling spatio-temporal data has been widely recognised as a powerful tool that captures far more of the meaning of concepts within a problem domain (Rojas-Vega and Kemp, 1994; Milne, Milton and Smith, 1993; Worboys, Hearnshaw and Maguire, 1990). It enhances the level of abstraction in a way close to our perception of the real world, offering a mechanism for expressing our understanding of the knowledge domain. Jackson (1994) advocates the use of object-oriented modelling in regional science as a common framework for integrating different semantics defined within social models. Object orientation is presented as a systematic approach to modelling the conceptual descriptors of complex socioeconomic models. It provides a way to formalise the handling of problems that need to be solved by the combined efforts of several people.

Bian (1997) has used the object-oriented paradigm to extend a two-dimensional static growth model into a three-dimensional dynamic framework. The aim was to study individual fish behaviour in an aquatic environment. In his object-oriented salmon growth system, the movement of individual salmon in a three-dimensional space was incorporated with the growth model to simulate the behaviour of salmon in selecting their habitat and their consequent growth. A number of simulations were run with five to ten adult salmon at a time for a period of several days.

However, the complexity of integrating object-oriented and geographic concepts into a spatio-temporal data model is an interesting challenge in its conception and its implementation. Choosing an object-oriented method is a laborious task. Object-oriented methods have been introduced into several distinct structures and representations, with over 50 published suggestions. ‘They range from the complex and difficult notations of OMT, Ptech and Shlaer/Mellor to the simpler ones of CRC and Coad/Yourdon, from an emphasis on process to an emphasis on representation and from language dependence to the giddiest heights of abstraction... . None of these methods is complete in the sense that all issues of the software development life cycle are addressed or that every conceivable system can be easily described’ (Graham, 1994, p. 287).

This book summarises a significant amount of research carried out in object orientation. Many of the concepts and implementations developed in this area are discussed and brought together within the context of GIS. The objective is to provide readers with a solid understanding of the object-oriented paradigm for designing a spatio-temporal data model.

1.2SPATIO-TEMPORAL DATA IN GIS

Representing spatial data in a GIS has been achieved by defining entities in geometric space in an explicit manner (vector representation) or an implicit manner (raster representation); see Burrough (1986). In the vector representation, three main geometric elements are used: points, lines and polygons, which are sets of vectors with interconnected coordinates linked to given attributes. The relationship among elements is represented by the connectivity of a set of vectors at the time of their storage into a GIS. For example, a set of lines is represented by starting and ending points, and some form of connectivity (straight line, curve, etc.). In a raster representation, entities are sets of cells located by their corresponding coordinates. In this case each cell is linked to an attribute value. The location of each cell is used to determine the adjacency relationship between entities.

As Dutton (1987) points out, the debate on vector versus raster representations is nearly as old as the concept of GIS. Both representations of geographic space have been regarded as valid data models. Besides, data transformation algorithms to convert from one spatial representation to another have been developed, and the choice between them is taken by the user who selects the representation that is most efficient for implementing a particular application in a GIS. Consequently, GIS has fully developed into information systems that are characterised by capabilities for representing, querying and manipulating entities in space. Over the past decade, expectations about exploring spatio-temporal data in GIS have raised interest in a wider range of capabilities. Some of these capabilities can be described as update procedures that are coherent with previous stored data, version management mechanisms to track the lineage of data, and analytical tools to recognise patterns of change through time as well as to predict future changes.

Representing spatio-temporal data in a GIS has been regarded as implementing an additional dimension in a former spatial representation (vector or raster). The primary objective for most of the spatio-temporal representations is summed up in the idea organising space over time. A geographic space is organised into partitions (layers) and the entities that inhabit this space are embedded in these partitions. In fact, a partition serves as a skeleton for representing several entities located in the geographic space at a particular point in time. This is a region-to-entity representation: first choose a region of a geographic space, then identify and locate the entities that inhabit that region according to how alike they are or how they are composed. Space and time dimensions are incorporated by determining their singularity through their contents; for example, space by attributes and shapes of the elements (points, polygons, lines, grid cells) and time by succession of happenings (events, actions, change, motion) on these elements. So far, this approach has been used in GIS by making spatially depicted classifications grouped into layers or sets of themes (e.g. geology, hydrology and land cover) between points or periods of time. In other words, geographic space is grouped along the spatial dimension after some sort of categorisation, and time is grouped along the time dimension after some sort of periodisation. Constituting history is explained based on similarity or dissimilarity between aggregations (layers) at different points of time (Figure 1.1).

Figure 1.1 Spatio-temporal layers as the main representation being used in GIS
(Reprinted with permission from Laurini and Thompson 1992, Academic Press Ltd)

Although this four-dimensional representation is sufficiently homogeneous for capturing and storing spatio-temporal data in GIS, it does not provide a unified representation of the real-world. We are dealing with geographic space: a space that reflects our knowledge of the environment where time exerts its influence on place in terms of human tasks and lived experiences. If we could decide, once and for all, which real-world phenomena should be represented as entities, relations or attributes in a geographic layer, our modelling task would be extremely simplified. In fact, what we need is to understand the nature of time itself with respect to the real-world phenomenon that we are trying to represent in a GIS. In order to accomplish that, the emphasis must shift from organising space over time to representing a real-world phenomenon in space and time.

This representation gives us an entirely different perspective to how we handle spatio-temporal data in GIS. It attempts to capture the complexity of space and time at the level of an indivisible unit - the entity. Instead of creating layers or time periods, this representation deals with elements’ coexistence, connection or togetherness. We are distinguishing two important concepts that are often regarded as interchangeable, an ‘entity’ and an ‘entity embedded in space’. This distinction would be unnecessary if we could always define the precise location of entities and their corresponding classified layers or time periods. In fact, we are confronted with a rather different reality. Most likely is that we may be uncertain of their location and how they change or move in a dynamic way. Moreover, we may know the location of an entity in a geographic space but we are uncertain of how to classify it. The notion of having an entity unconstrained by its surroundings in space and time allows us to examine how a real-world phenomenon is represented independently of how geographic space is organised at a particular time.

This is a space-time entity representation: first identify the entities, and second ensure that based on these entities a geographic space can be created. An important characteristic of this representation is the ability to create the geographic space based on a specific task to be solved or a particular knowledge about the real-world at a particular point in time. Depending on the specific task to be solved or the human ability to see the world at a particular point in time, certain real-world phenomena may be represented as entities in a geographic space, and others become the relations we are interested in modelling. For other tasks or different perspectives in the world, these roles may change. Therefore, modelling spatio-temporal data in GIS becomes an exercise of understanding not only the similarities and dissimilarities between regions of geographic space, but also the coexistence (connection or togetherness) relationships between the entities that inhabit these regions.

A reliable space-time entity representation is needed when designing a spatio-temporal data model in GIS. As Peuquet points out, a variety of approaches for studying space-time phenomena has evolved in social, geographical and physical studies. ‘Andrew Clarks’s early work on historical geography demonstrated that changing spatial patterns could be studied as “geographical change” (Clark, 1959, 1962). Cliff and Ord (1981) later examined change through time by scanning a sequence of maps, searching for systematic autocorrelation structures in space-time in order to specify “active” and “interactive” processes. Perhaps the best-known efforts within the field of geography that made explicit use of time as a variable in the study of spatial processes are Hägerstrand’s models of diffusion and time geography’ (Peuquet, 1994, p. 441).

1.3TIME GEOGRAPHY

Torsten Hägerstrand, a Swedish geographer, unfolded the Time Geography approach in the early 1960s. He examined space and time within a general equilibrium framework, in which it is assumed that every entity performs multiple roles; it is also implicitly admitted that location in space cannot effectively be separated from the flow of time. In this framework, an entity follows a space-time path, starting at the point of birth and ending at the point of death. Such a path can be depicted over space and time by collapsing both spatial and temporal dimensions into a space-time path. Time and space are seen as inseparable.

Time Geography has provided a foundation for recognising paths of entities through space and time and for uncovering potential spatio-temporal relationships among them. Moreover, its application in various areas has produced the concept of a ‘continuous path’ to represent the experience occurring during the lifespan of an entity. This experience is in fact conceptualised as a succession of changes of locations and events over a space-time path. Most of the applications using Time Geography have been devoted to modelling individual activity paths within a period of time, analysing the pattern of activities for any individual path, as well as simulating individual activity paths.

This book proposes a new means for applying the time geography approach. Its goal is to employ the concept of a space-time path developed in time geography for representing spatio-temporal data within a spatio-temporal data model. The time geogra...