Julie Sungsoon Hwang & Jean-Claude Thill Department of Geography State University of New York at Buffalo U.S.A. August 24, th Int’l Symposium of Spatial Data Handling Empirical study on location indeterminacy of localities
Research question How can we represent vague concepts of spatial object in a (discrete) computing environment (e.g. GIS)? Nearness in localities Mental maps of localities Indeterminate boundaries of localities
Research scope Mental maps Generals: f (distance, relation, scale) Specifics : f (preferences, experience, …) Localities Official recognition: eg. administrative unit Unofficial recognition: eg. vernacular region
Research objective [1] Building the model of locality boundary using fuzzy regions (egg-yolk model) and some rules regarding nearness 0 1 BA A B 2-Dimensional Geographic Space x: 1-Dimensional Geographic Space Y: Degree of Membership
Research objective [2] Examining any difference in location indeterminacy between urban and rural settings Buffalo Urban Wilson Rural
Example: identifying localities… Accident location? Which city?
Task 1: theoretical Building the model of locality boundary using fuzzy region and rules of nearness
Fuzzy regions Core Exterior Boundary
Nearness = Fuzzy set membership of belonging to “Syracuse” Near “Syracuse”? What determines the fuzzy set membership function value? What determines the fuzzy set membership function value? Euclidean distance Euclidean distance Spatial qualitative relation Spatial qualitative relation Scale-dependent Scale-dependent
Locality as a fuzzy region Exterior Core Boundary 1stOrderGr 2ndOrderGr
Computing fuzzy set membership value in GIS: work steps 1. Delineate boundaries 2. Assign membership values 3. Create TINs4. Interpolate values on TINs
Computing fuzzy set membership value in GIS: results
Comparison to other proximity measures coreexterior0.5-cut boundary coreexterior0.5-cut boundary core Distance BufferFuzzy proximity
Task 2: empirical Examining any difference in location indeterminacy between urban and rural settings
Georeferencing traffic accident data We considered 5460 out of 8631 cases from NYS ‘96-’01 Of these, 246 urban, and 298 rural localities are compared
Computing location indeterminacy index of localities i = 1 - (Σ i )/n 78% sure95% sure58% sure
Comparing location indeterminacy index of urban versus rural localities Average number of fatal crashes in rural areas is 2 whereas those in urban areas is 16 To work around small number problem, we compute Bayesian estimates of both groups adjusted for within-group distributions People are 94% (or somewhere between 93% and 95%) sure in identifying urban localities while they are 88% (or somewhere between 86% and 90%) sure in identifying rural localities
ANOVA Analysis of variance conducted on Bayesian estimates of location indeterminacy confirms the difference between urban versus rural locality is significant in terms of location indeterminacy Neighborhood types may affect the degree of certainty to which the boundary of locality is perceived
Interpretation of results Mental maps of urban settings may be less error-prone than those of rural settings Spatial knowledge acquisition: city provides more landmark or route upon which judgment on indeterminate boundaries of localities can be based Scale factor: dense urban settings provide a reasonable scale in which humans can conceptualize localities without much difficulty
Conclusions Fuzzy set theory provides a reasonable mechanics to represent vague concept of geospatial objects Neighborhood types affect the way humans acquire spatial knowledge and forge mental representations of it