Raster data represents the world as a regular grid of cells (pixels), each storing a single value for a defined location. It is one of the two fundamental GIS data models, the other being vector. Rasters are ideal for continuous phenomena: elevation, temperature, slope, satellite reflectance, and classified surfaces such as land cover.
Why it matters
Terrain and remote-sensing work is largely raster work. A digital elevation model (DEM) is a raster of elevation values; slope, aspect, hillshade, and flow accumulation are all rasters derived from it cell by cell. Cell size (resolution) sets the level of detail and directly affects analysis results: a 30 m DEM cannot resolve features a 1 m LiDAR DEM can, and resampling between resolutions changes derived values.
Concrete example
A Sentinel-2 image is multiband raster data: band 4 (red, 10 m), band 8 (near-infrared, 10 m), and band 11 (SWIR, 20 m) are each grids of reflectance values. NDVI is computed per cell as (NIR − Red) / (NIR + Red). A GeoTIFF (often a Cloud-Optimized GeoTIFF) is the standard container, storing the grid plus georeferencing (origin, cell size, CRS) and a NoData value for empty cells.
Common pitfall
Computing slope or area on a DEM stored in geographic coordinates (EPSG:4326): cell sizes are in degrees, not metres, so a degree of longitude shrinks toward the poles and slope values come out wrong. Reproject to a metric CRS first. Also watch NoData handling at tile edges, which can create false flat areas or artifacts.