Raster data model

These are the basic properties of the raster data model. Raster data are regular grids (matrices) made up from individual cells with some arbitrary values describing something. The values are only limited by the type of the storage. They can be in the range of bytes, 8-bit integers, 16-bit integers, floating point numbers, and so on. Rasters are always rectangular (like an image); however, they can give a feeling of having some other shape with a special kind of value: NULL or No-Data.

One of the most useful properties of raster data is that their coverage is continuous, while their data can change. They cover their entire extent with coincident cells. If we need a full and continuous coverage (that is, we need a value for every point describing a dynamically changing phenomenon), raster is an obvious choice. On the other hand, they have a fixed layout inherited from their resolution (the size of each cell). There are the following two implications from this property:

• First, the accuracy of a raster is not constant. It covers uniform areas in a given projection. Therefore, on the globe, the area covered by a single raster inherits the distortion of the projection.
• Secondly, if we increase the resolution, the size of the raster data shows a quadratic growth as we have to increase the number of cells in each dimension:

Of course, this property works in two ways. Rasters (especially with square cells) are generally easy and fast to visualize as they can be displayed as regular images. To make the visualization process even faster, QGIS builds pyramids from the opened rasters.

Using pyramids is a computer graphics technique adopted by GIS. Pyramids are downsampled (lower resolution) versions of the original raster layer stored in memory, and are built for various resolutions. By creating these pyramids in advance, QGIS can skip most of the resampling process on lower resolutions (zoom levels), which is the most time-consuming task in drawing rasters.

The last important property of a raster layer is its origin. As raster data behaves as two-dimensional matrices, it can be spatially referenced with only a pair of coordinates. These coordinates, unlike in graphics, are the lower left ones of the data. Let's see what QGIS can tell us about our raster layer. We can see its metadata by right-clicking on it in the Layers Panel, choosing Properties, and clicking on the Metadata tab:

As you can see in the preceding screenshot, our raster layer has a number of rows and columns, one band, an Origin, a resolution (Pixel Size), a No-Data value, and a Data Type.

To sum up, the raster data model offers continuous coverage for a given extent with dynamically changing, discrete values in the form of a matrix. We can easily do matrix operations on rasters, but we can also convert them to vectors if it is a better fit for the analysis. The raster data model is mainly used when the type of the data desires it (for example, mapping continuous data, like elevation or terrain, weather, or temperature) and when it is the appropriate model for the measuring instrument (for example, aerial or satellite imaging).