Spatial data represents geographic features using two fundamental models. The vector model represents discrete features as points, lines, and polygons defined by coordinate pairs, with associated attribute tables -- ideal for boundaries, networks, and discrete objects (buildings, roads, parcels). The raster model represents continuous phenomena as regular grids of cells (pixels), each storing a value -- ideal for elevation, temperature, satellite imagery, and any field that varies continuously across space. Choosing between vector and raster (or combining both) depends on the nature of the phenomenon, the analysis to be performed, and the required precision of boundary representation.
The choice of spatial data model is the first and most consequential decision in any GIS project, because it determines what analyses are possible, how storage and processing scale, and how accurately the real world is represented.
The vector model excels at representing discrete features with well-defined boundaries. A land parcel is naturally a polygon with precise coordinates; a road is naturally a line; a fire hydrant is naturally a point. Each feature links to a row in an attribute table, making vector data ideal for database-style queries ("select all parcels zoned commercial with area over 1 hectare"). Common vector formats include Shapefile, GeoJSON, GeoPackage, and features stored in spatial databases (PostGIS, SpatiaLite).
The raster model excels at representing continuous phenomena -- elevation, temperature, precipitation, satellite imagery, land cover probability. Each cell in the grid stores a single value, and the grid's resolution (cell size) determines the spatial detail. Raster data is computationally efficient for map algebra -- cell-by-cell mathematical operations like adding two layers, computing slope from elevation, or creating NDVI from red and NIR bands. Common raster formats include GeoTIFF, NetCDF, and cloud-optimized GeoTIFF (COG).
Many analyses combine both models. A flood analysis might use a raster DEM for hydrological modeling, then convert the inundation boundary to a vector polygon for overlay with vector parcel data to identify affected properties. Remote sensing classification produces raster land cover maps that are often vectorized for integration with administrative boundaries. Understanding the strengths, limitations, and conversion pathways between vector and raster is essential for effective GIS work.