Geographic Information Systems
GIS questions test data models, topology and whether you can predict the output of an operation before running it.
Data models
- Vector data uses points, lines and polygons with explicit topology (adjacency, connectivity, containment). Topology prevents slivers and gaps and enables network tracing.
- Raster data is a regular grid with one value per cell; cell size is the fundamental accuracy limit, and mixed pixels cause error along boundaries.
- Object-relational geodatabases add domains, subtypes and relationship classes above the relational model.
Coordinate handling
Layers in different coordinate systems display together only through transformation; the stored data is unchanged. Reprojection recomputes vector geometry and resamples rasters, and is not lossless.
Operations
Overlay tools (clip, intersect, union, erase, identity) combine geometry and attributes. The number of output features depends on the geometry of both inputs, not on attribute counts. Spatial join transfers attributes by spatial relation; select-by-location is the query form of the same idea.
Worked example: a 10 m raster is resampled to 30 m. Cell area grows by (30/10)² = 9, so the cell count falls to one-ninth. Nearest-neighbour resampling preserves class values and is correct for land cover; bilinear or cubic interpolation creates new values and corrupts categorical data. Separately, clipping buildings to a flood polygon keeps only building attributes, whereas intersecting keeps attributes from both layers - a distinction examiners test directly.
Traps
- Assuming clip and intersect produce equivalent attribute tables.
- Storing area or length as an attribute and then editing geometry without recomputing it.
- Applying interpolating resampling to categorical rasters.