First, an often-touted strength of GIS involves the linking of spatially referenced data sets, many times collected by different agencies for different purposes. Assessments of data quality and accurate analysis (and visualization) of data uncertainty are critical to attaching scientific inference to data linked and displayed in a GIS. The accurate use of GIS in spatial analysis requires development of tools for addressing uncertainty (both statistical and deterministic) within the GIS environment. Otherwise, GIS offers many advances in creative data display and management, but actual analysis of data primarily occurs outside of the GIS (either in the head of the viewer of GIS output, or in specialized software packages). In short, without accurate presentation of uncertainty, associations between variables displayed in a good map of bad data often appear more believable than those displayed in a bad map of good data, and there is sizable potential for misinterpretation.
Second, many application areas focus on the analysis of observational
rather than experimental data. Different aspects of key issues in the analysis
of observational data surface in different application areas. For example,
the "modifiable areal unit problem" of geography has aspects in common
with sociology's and epidemiology's "ecologic fallacy" of assigning associations
observed in aggregate to individuals. Also, the latent variables of econometrics
and the notion of
unmeasured confounding in epidemiology are differently named manifestations
of the same problem. While these issues are not necessarily synonyms, they
nonetheless reflect different facets of deeper issues underlying all analyses
of observational data. As a result, there is need for more interdisciplinary
collaborations providing fundamental advances in spatial analysis of observational
data without "reinventing (or renaming) the wheel".
Advances in the utility of GIS and spatial analysis in any of a variety of application areas similarly would profit from interdisciplinary developments. GIS advocates the introduction of spatial thinking into application areas, however I also believe GIS can benefit from the introduction of application area thinking. For example, the field of epidemiology can benefit from the introduction of spatial analysis techniques, but not at the sacrifice of well-developed concepts such as confounding, causation, and the ecologic fallacy. The underlying goal in introducing GIS as an analysis tool should not be only to enable novel analyses, but to enable better analyses. In some cases this will involve construction of new analytic techniques enabled by GIS, but in others the true utility of GIS lies in increased efficiency in design and implementation of established approaches. Determining which is most appropriate requires creative insight from GIS developers, spatial analysts, and application area experts.