Geography Reference
In-Depth Information
FIGURE 4-6 The quadtree
technique for storage of raster
The computation of “where” becomes more complicated if the matrix of cells is not oriented
along Cartesian grid lines or the graticule formed by meridians and parallels, but, after all,
computation is something that computers are good at. A more subtle complication occurs if the area
covered by the raster is large in a north-south direction because of the issues related to projecting the
curved earth onto a flat plane. However, for the most part, the matter of location using a raster
approach is easily handled.
If a raster cell contains more than one type of area, as many on boundaries between areas do, there is
an approach called “quad tree” in which the raster cell is divided into four subcells, as in Figure 4-6.
If a subcell is homogeneous in the feature value, then it is left alone. If not, it is redivided into four
more sub-subcells, and the process is repeated. This redivision continues until all the subcells are
homogeneous, or they become too small to make further subdivisions reasonable.
Vector Data Model
Using the vector approach, the level of difficulty of determining “what” and “where” is somewhat
reversed. The “what” is relatively simple. Each unique area is enclosed inside a polygon, so the content of
the polygon is homogeneous—containing a unique value (or a unique set of values) related to the theme.
Contrast this with the raster approach, where several features or conditions may occur in one cell and
(usually) one feature or condition is picked for recording in the database.
The “where” with the vector approach is a bit more problematical but usually allows greater precision.
Whereas with a raster approach, the location of each area (cell) is but a simple calculation, with the
vector approach, each vertex of each polygon has an explicit geographic location. Therefore, the “where”
question may have thousands to millions of answers, in terms of coordinate pairs for a large area or
one with many polygons that have complex boundaries. Of course, the issue with respect to projection,
datum, and so on are present—for all the coordinate pairs.
 
Search WWH ::




Custom Search