Modelling spatio-temporal environmental data

A conceptual model for environmental data is presented with special emphasis on the ability to store spatio-temporal references of the data. Other aspects of the model are the ability to handle hierarchical data and semantics of the measurements. Introduction How people perceive things and their relationships is not only the basis of forming theories but also for collection and storage of information. The reality with its immense complexity needs to be abstracted somehow in order to be manageable. The first step of modelling in connection of data management, conceptual modelling, does just that. Its goal is to produce a model that describes how the users perceive the data; what kind of entities does the data consist of, and what are their relationships (Elmasri and Navathe 2000). Most of the current conceptual models for environmental data are atemporal in the sense that they only model the current status of their microworld. This is reasonable as up-to-date information is required for most uses of environmental data like the decisions of possible management actions. Moreover the future development can usually be projected from the current status well enough for the purpose of operational planning; an example would be a forestry database where information content is utilized in predicting the development of forests to time operations like planting, thinning and harvesting (Paananen 1994, Tokola et al. 1997). Some planning tasks would, however, be facilitated if the history of the forest would be known instead of only the current situation. In the recent years more and more emphasis is put globally on the changes on environment and biodiversity in particular. For the assessment of these changes to be feasible, information about the history is required. What have been the past states of environment, when and how have they changed? These questions set new demands on the way environmental information is collected and stored. Model as an abstraction of reality What are the entities under study and how the information concerning them is obtained bears significant relevance to the way the entities are modelled. Particularly this is true for the spatial properties of environmental data. Molenaar (1998) list two spatio-temporal processes: one with field characteristics meaning position dependent field values, and another one based on the behaviour and state of objects. These result respectively in continuous and discrete views of the geographic space. But he also states that "we have to realise that computers are finite state machines that work on discrete data". Thus by definition also the field processes are approximated by discrete representation in data management. A common situation with environmental data is that it is modelled as a tessellation of the earth's surface. This tessellation can be either regular giving relevance only to location and the value associated with it or it may be irregular emphasizing the identity of the modelled feature. In this model the abstraction of physical reality is founded on features. The basic unit of modelling is ST_feature which has as an attribute its geometry and location (figure 1). An instance of ST_feature has a lifetime expressed as an attribute valid_period . During its lifetime it may have several geometries and locations, again valid for certain periods during the lifetime. Even though the principal approach to spatial modelling is based on identifiable features, it is quite easy to see that the model can equally produce a discrete approximation of continuous field. In that case the shapePart would consist either of single points or regular tessellation units (grid cell) which are constant throughout the timeframe of the database. 5 AGILE Conference on Geographic Information Science, Palma (Balearic Islands, Spain) April 25-27 th 2002