A raster data model typically represents geographic features (e.g., land cover) or spatial phenomena (e.g., ocean surface temperature) using a matrix of uniformly sized square areas. Each square, which may be referred to as a cell, grid, or pixel (picture element) stores a single value that represents what is being mapped. If the raster dataset represents ocean surface temperature (continuous data), then each cell value would record the average temperature for the geographic area the cell covers; if the raster dataset represents land cover (categorical data), then the cell value could be an integer that represents a predominant landscape class such as forest or water that the cell covers.

Each cell represents a physical square on the Earth's surface. The ground size of cells (i.e., the length of each side of the cell in map units) affects the detail that can be captured. Raster datasets that use a large cell size are generalized since whatever variability exits within the area covered by the cell must be reduced to a single value. Small cell sizes capture more variability and therefore more detail. Small cell sizes also mean more data and thus larger file sizes and possibly higher costs (i.e., high resolution imagery).

Raster data formats are particularly well suited for encoding data, such as elevation, precipitation, temperature, and wind speed that vary continuously over geographic space and are commonly used with satellite imagery, aerial photography, or scanned documents such as platmaps, deeds, and USGS topographic maps.

Vector data, on the other hand, represent geographic features and spatial phenomena as points, lines, or polygons (areas). A major difference between raster and vector data is that raster data represent geography as the interior area of a cell, whereas vector data encode the boundaries between different areas.