|
Abstract: Many of us have a considerable depth of process knowledge and understanding in the specific domains of our specialisations in addition to the skills we have in GISc. Those of use whose specialist domain is GISc have collaborators with specialisations in other, useful, domains. Only very rarely is that process knowledge and understanding effectively brought into play when we are analysing spatial data. The gains when it is used effectively are considerable.
As most of our analytical techniques are concerned with partitioning, separability or clustering, then the spatial relationships within the data domain chosen for an analysis are fundamental to the success, or otherwise, of that analysis.
Data can be envisaged as being distributed in a number of domains simultaneously. Changing domains allows one to structure the data more effectively for analysis or modelling. Putting it more formally, the topology changes from domain to domain. Take, for example, a natural resources task. The data can be envisaged in Geographic Space (in a GIS or on a map), in Spectral Space (in an image analysis package), or in Environmental Data Space (in a package such as ISOCLIM or BIOCLIM).
The data behaves differently in each of these data spaces. Each location in Geographic Space may occupy only one location in Spectral or Environmental data space. A point in either Spectral or Environmental data space may occupy one or more locations in Geographic Space. A large number of points in Geographic Space may occupy the same positions in Spectral or Environmental data space. A problem which is intractable in one data space may be easily solved in an alternative data space. This presentation explores this aspect of data analysis with several examples and looks at other ways of introducing process understanding into GIS modelling.
|
