GIS and Spatial Analysis in Urban Planning
This domain tests data models, coordinate reference systems, standard spatial operations and the interpretation of spatial statistics. Expect a question that gives a statistic and asks for the correct reading.
Rules
- Vector vs raster - discrete features with topology versus continuous surfaces; buffers and overlays on vector data; suitability modelling by weighted raster overlay.
- CRS - WGS84 geographic coordinates versus projected UTM; Iran spans UTM zones 38N to 41N, so a national dataset must be projected consistently before measuring area or distance.
- Operations - clip, intersect, union, spatial join, buffer, dissolve, network analysis, geocoding, kernel density.
- Spatial statistics - Global Moran's I for clustering, Getis-Ord Gi* for hot spots, nearest-neighbour index, 2SFCA for accessibility; the modifiable areal unit problem (MAUP) and the ecological fallacy constrain inference.
Worked example
A parcel-level analysis of building permits returns Moran's I = 0.35 with z = 2.4 and p = 0.02. Reading: positive spatial autocorrelation - high-permit parcels sit near other high-permit parcels, and the result is unlikely under spatial randomness. It does not say where the clusters are; that needs Gi*. It also does not establish why - a corridor or a single large developer could produce the pattern. Finally, if the study were re-run on district aggregates rather than parcels, the value could change purely from MAUP. Reasoning shown: direction (positive), significance (p < 0.05), localisation (absent), and the inference limit (no causality).
Traps
- Reading p as effect size; a large sample makes a trivial cluster significant.
- Using straight-line buffers to represent network access.
- Mixing datums and then measuring areas.