Geography Reference
In-Depth Information
FIGURE 8-8 Raster representation creates jagged
lines and polygons
siting problem. Raster methods may suffer, when tightly zoomed in during display, from what is
inelegantly called the “jaggies” (see Figure 8-8), where polygonal and linear features are shown in
stair-step fashion. This actually may not be a bad thing in many cases, since the precise lines separating
polygons usually imply a level of accuracy that does not exist. In human-made phenomena (like parcel
boundaries or building footprints), there are such precise lines, but they may not be represented in
exactly the right place. In natural phenomena (e.g., the edge of a forest), there is usually little chance that
the precise one-dimensional line accurately represents an actual boundary. Here the jagged boundary is a
nice reminder that we don't know precisely where the boundary is—at least partially because there isn't
a precise boundary.
A main objection to using raster methods on the kinds of problems we have used vector methods to solve
is that the minimum units for line and point representation all have to be the same size. So, for purposes
of overlaying, a tree and a highway usually wind up being the same width. But, as has often been stated
in this text, GIS only serves up an approximation of the real world, and the proof of its utility is in
whether it promotes comprehension of the real world, not whether it mimics the real world.
Another problem is that one size does not fit all. You probably wouldn't want to store moose habitat at
the same cell size as soil type. Adjustments can be made during analysis—generally toward using the
largest cell size—but again there are compromises.
Another use of rasters—an important use, mentioned before: A raster can pretend it is a surface. A surface
is a mathematical construct such that, for every value of x and y on a horizontal Cartesian plane, there
is a single z value that indicates a displacement from that plane. Any change in x and/or y, no matter
how tiny, may result in a change in the z value. In pretending to be a surface, a raster cell clumps a whole
bunch (an infinite number) of x-y points together and gives them a single z value. The surface is therefore
lumpy. Thus, like much of GIS, it provides an approximation to the hypothetical (mathematical) world,
which itself is endeavoring to provide an approximation of the real world.
Rasters can be effectively used in simulations. Given a raster of a forest fire covering a certain area, you
could plug in other rasters that indicated wind speed and direction, slope, concentrations of fuel, and
so on. From this you could make predictions about where the fire would go next. The fact that each cell
relates geographically to its neighbors in only one of two ways (side by side, or diagonally) provides
tremendous modeling and computation advantages.
Search WWH ::
Custom Search