Object-based analysis to map latent spatial phenomena

The design of strategies to adequately represent complex geographical phenomena is a core objective of the GIScience discipline.  Geons are domain-specific units, semi-automatically delineated with expert knowledge incorporated, scaled and of uniform response to a phenomenon under concern. The average size, composition and internal homogeneity of geons are controllable by logical models that operationalize the complex phenomenon and are fed into the regionalization process. These models specify (1) the multi-dimensional set of variables (or indicators) representing the phenomenon under concern, (2) the definition of internal homogeneity (functional composition, dimensional weights, etc.), (3) the scale of intervention measures. We follow a systemic stability concept by minimizing entropy (heterogeneity) within the generated units at an optimized information scale.

The project builds upon ongoing research on the spatial implications of indicator selection and weighting, local and global statistical robustness, and the intervention scale. The project shall bridge principles and conceptual findings from the composite indicator community with technical achievements of the OBIA paradigm, to optimize spatial composite indicators. Such meta-indicators should be likewise methodologically sound, statistically robust, salient in meaning and valid for those who use it. A challenge is to transform composite indicator values into nominal categories, reflecting (a) internal composition as well as (b) spatial and relational characteristics.

Research questions include:

1.       Is scale-specificity adequately reflected in established autocorrelation / entropy measures?

 

2.       Which strategies exist to deal with the non-conformity and statistical ‘flaws’ of various input data sets?

 

3.       Can we assign labels to the generated units and if yes, is there a way to establish a transferable taxonomy for domain-specific geon sets?