Using geospatial analysis we try to describe, explain and predict geographical phenomena. Theories and methods adopted from mathematics, statistics, computer graphics and information theory have been integrated with GI Science approaches to yield a mature and useful toolbox for such analysis.
Spatial statistics represents one of the most important and core methodologies. Although not a new area in GI science, there is a scope to expand its applications considerably. In spatial data mining it is one of the core techniques and specific topics such as geostatistics, spatial autoregressive processes and point processes deliver techniques of considerable interest. When applied to multivariate analysis, further specialist methods such as traditional principal components analysis and factor analysis, or more recent self-organizing maps and k-means clustering analysis can be used.
The development of realistic geospatial process models and those which incorporate time (spatio-temporal models) in a realistic manner will lead to improved representations of the real world. The models themselves must be understandable, applicable to a range of data sets and situations and must be capable of integration with others in processing workflows: geospatial process ontology needs development to ensure compatibility and interoperability.
Several computational methods can be used in implementing these geospatial modelling and analysis methods. Intelligent agents, cellular automata, neural networks and fuzzy logic are examples of geocomputational methods, which have not yet been adopted as standard computational solutions in GI applications. Algorithm development is often undertaken on an ad hoc basis for specific tasks, but it may well use particular spatial data structures, such as Voronoi and TIN models, or use particular approaches, such as data compression (e.g. wavelets) or network analysis based on graph theory. The latter, in particular its extensions (e.g. labelling and weighting of graphs), has not been researched and applied enough in spatial problem solving.
All these techniques to get spatial information and create spatial knowledge, related to data quality and risk issues, can be implemented to support spatial decision-making.