Spatial Analysis and Geoprocessing Methods
This domain tests whether you can choose and sequence the right operations to answer a spatial question: overlay, proximity, surface and terrain analysis, network analysis, raster algebra and spatial statistics.
Overlay and proximity
- Intersect keeps the overlap of two layers; union keeps everything; erase removes one layer's area; identity combines both while retaining all input features.
- Buffers can be fixed distance, variable distance from a field, or ring (multiring) buffers; dissolving removes internal boundaries. In dense data, dissolved buffers prevent double-counting overlapping catchments.
- Distance surfaces and near tables answer "how far", allocation answers "which is closest".
Surfaces and terrain
Slope, aspect and hillshade derive from a DEM and are only as good as its vertical accuracy and resolution. Viewshed analysis needs an observer offset and earth-curvature correction for long distances. Hydrology tools run in sequence: fill sinks, flow direction, flow accumulation, stream definition, stream link, watershed delineation. Skipping the fill step produces broken flow lines.
Raster algebra and statistics
Map algebra combines rasters cell by cell; zonal statistics summarises a value raster within zones; reclassification converts continuous values into ranked classes for suitability modelling. Spatial statistics include Moran's I for global clustering and Gi* for local hot spots, both requiring a spatial weights definition.
Worked example
A suitability model weights three criteria: slope (0.5), distance to road (0.3) and land cover (0.2). Each criterion is reclassified to 1-5, where 5 is most suitable. A cell has slope score 3, road score 5, land cover score 2. Weighted score = 0.5(3) + 0.3(5) + 0.2(2) = 1.5 + 1.5 + 0.4 = 3.4. Reasoning: weights must sum to 1, all inputs must share the same cell size, extent and coordinate system, and reclassification breaks must be justified rather than arbitrary - otherwise the score is arithmetic without meaning.
Traps that cost marks
- Ignoring the modifiable areal unit problem: changing zone boundaries changes results even with identical underlying data.
- Drawing individual conclusions from aggregate data (ecological fallacy).
- Forgetting that kernel density output is intensity per unit area, not a count.
- Running overlay on layers with mismatched coordinate systems or extents, silently dropping features.