Spatial Analysis and Modeling
Spatial analysis questions test whether you can match a method to the question and defend your parameters.
Analysis families
- Overlay and buffers answer proximity and containment questions; buffers are Euclidean unless a network or cost surface is used.
- Network analysis: shortest path (Dijkstra), service areas, closest facility, origin–destination matrices; impedance may be distance, time or cost.
- Terrain: slope, aspect, hillshade, flow direction and accumulation, watershed delineation, viewshed.
- Interpolation: IDW is exact but produces bull's-eye artefacts; splines are smooth but can overshoot; kriging uses a fitted variogram, reports a variance surface, and ordinary kriging assumes a constant unknown mean.
- Suitability modelling reclassifies criteria to a common scale, weights them, then combines - weighted overlay or multiplicative.
Spatial statistics
Global Moran's I near +1 indicates clustering, near 0 randomness, near −1 dispersion; local indicators map where the clusters are. The ecological fallacy and the modifiable areal unit problem (MAUP) mean conclusions can change with the aggregation zone.
Worked example: two adjacent DEM cells, 10 m apart, have elevations 150.0 m and 153.0 m. Slope = 3.0 / 10 = 0.30, i.e. 30% or arctan(0.30) = 16.7°. Quoting "30" without units or without saying percent-versus-degrees loses the mark.
Traps
- Reporting degrees as percent, or vice versa.
- Using IDW on strongly trending data.
- Ignoring edge effects on slopes computed at the raster boundary.
- Presenting Moran's I without stating the spatial weights definition.