A raster’s coordinate system may be in real-world coordinates or “image space.” For those in geographic space, cells are referenced primarily by an (x,y) location in map coordinate space, not by row and column numbers, as previous versions of grid processing were. The location referenced for a given cell is the center of that cell. For the overall raster, the geographic reference is usually the center of the lower-left (southwest) cell. Cells are square in map coordinates.
Each cell represents a specific location on the surface of the Earth. While each cell of a raster is square, the surface area of the Earth the raster represents may not be. Such real-world areas should be almost square, however. If the northing span of a raster is too great, the representation suffers. For example, take a UTM zone whose “base” is at latitude 40 degrees and “top” is at latitude 48 degrees. Suppose the cells are 100 meters on a side. Let’s say a cell at the base would represent 10,000 square meters. A cell at the top of the zone, admittedly more than 500 miles away, would represent only about 8,800 square meters!
Rasters may be described as integer or floating-point. All the cells in an integer raster contain integers or, if not, are designated NoData. All the cells in a floating-point raster contain floating-point numbers, or, if not, are designated NoData.