Georeferenced Raster Data

Raster geodata consists of georeferenced data grids and images that are stored as matrices or objects in the MATLAB® workspace. While raster geodata looks like any other matrix of real numbers, what sets it apart is that it is georeferenced, either to the globe or to a specified map projection, so that each pixel of data occupies a known patch of territory on the planet.

All regular data grids require a referencing object, matrix, or vector, that specify the placement and resolution of the data set. Geolocated data grids do not require a separate referencing object, as they explicitly identify the geographic coordinates of all rows and columns. For details on geolocated grids, see Geolocated Data Grids.

Referencing Objects

A spatial referencing object encapsulates the relationship between a geographic or planar coordinate system and a system of intrinsic coordinates anchored to the columns and rows of a 2-D spatially referenced raster grid or image. A referencing object for raster data referenced to a geographic latitude-longitude system can be a GeographicCellsReference or GeographicPostingsReference object. A referencing object for raster data referenced to a planar (projected) map coordinate system can be a MapCellsReference or MapPostingsReference object. Unlike the older referencing matrix and vector representations (described below), a referencing object is self-documenting, providing a rich set of properties to describe both the intrinsic and extrinsic geometry. The use of referencing objects is preferred, but referencing matrices and vectors continue to be supported for the purpose of compatibility.

Referencing Matrices

A referencing matrix is a 3-by-2 matrix of doubles that describes the scaling, orientation, and placement of the data grid on the globe. For a given referencing matrix, R, one of the following relations holds between rows and columns and coordinates (depending on whether the grid is based on map coordinates or geographic coordinates, respectively):

[x y] = [row col 1] * R, or
[long lat] = [row col 1] * R

For additional details about and examples of using referencing matrices, see the reference page for makerefmat.

Referencing Vectors

In many instances (when the data grid or image is based on latitude and longitude and is aligned with the geographic graticule), a referencing matrix has more degrees of freedom than the data requires. In such cases, you may encounter a more compact representation, a three-element referencing vector. A referencing vector defines the pixel size and northwest origin for a regular, rectangular data grid:

refvec = [cells-per-degree north-lat west-lon]

In MAT-files, this variable is often called refvec (or maplegend). The first element, cells-per-degree, describes the angular extent of each grid cell (e.g., if each cell covers five degrees of latitude and longitude, cells-per-degree would be specified as 0.2). Note that if the latitude extent of cells differs from their longitude extent you cannot use a referencing vector, and instead must specify a referencing object or matrix. The second element, north-lat, specifies the northern limit of the data grid (as a latitude), and the third element, west-lon, specifies the western extent of the data grid (as a longitude). In other words, north-lat, west-lon is the northwest corner of the data grid. Note, however, that cell (1,1) is always in the southwest corner of the grid. This need not be the case for grids or images described by referencing objects or matrices.