the width and left- and right-hand address
ranges of a street segment.
Each format has particular advantages.
Maps in raster format, for instance, are well
suited for print-on-demand distribution as
well as dissemination over the internet
(Evans and Vickers, 1995). By contrast, vector
data are especially useful for creating map dis-
plays at different scales on diverse projections
(seemap projection). Mathematical formulae
afford ready conversion between spherical and
plane coordinates or from one projection to
another. Indeed, tailored map projections that
control distortion for a region of interest or
provide insightful views, including dynamic
oblique views or flyovers, are a prime asset of
digital cartography.
Digital cartographic research has also pro-
ducedalgorithmsfor map generalization and
map labelling, both of which support the
dynamic, interactive change of scale for street
maps, topographic maps and electronic
atlases. Generalization is especially relevant
for maps displayed atscalessmaller than that
for which the data were originally acquired
(McMaster and Shea, 1992). Because map
symbols are more likely to overlap as scale
decreases, a generalization algorithm must
first identify areas where overlap will occur
and then avoid aesthetically awkward (and
sometime confusing) graphic conflict by sup-
pressing less important features or shifting
them apart – one of several ethically respon-
sible trade-offs that human mapmakers
call ‘cartographic license’ (Li and Su, 1995).
Generalization must also smooth out intricate
coastlines and streams with tight meander
loops, both of which can yield ambiguous
and unappealing blobs at substantially reduced
scales. Automated map labeling is similar to
map generalization insofar as the algorithm
must determine potential conflict between
the allegedly ideal positions of map labels
(Zoraster, 1997). If the conflict cannot be
resolved by moving one or more labels to a
less desirable location, the algorithm might
need to leave a feature unlabeled or use a thin
leader line to link a symbol with a label that
cannot be adjacent. More straightforward is
the attachment of text to curved features such
as roads and streams.
Another common display task in digital
cartography is the realistic viewing of digital
elevation models. Typical strategies include
low-angle oblique views, which require a geo-
metric transformation as well as the identifica-
tion and suppression of symbols where the
surface is hidden from view (De Floriani and
Magillo, 2003) and the addition of shadows,
or hill shading, in areas not illuminated by
a hypothetical light source located in the upper
left or upper right (Tuhkanen, 1987). More
intriguing are maps formed by draping satellite
imagery, land-cover classifications or topo-
graphic symbols over obliquely viewed digital
elevation models (Banerjee and Mitra, 2004).
Although digital cartography has largely
displaced its non-electronic counterpart as
the principal means of acquiring and storing
geospatial data, paper maps thrive in diverse
ways: in newspapers, magazines, atlases and
geography textbooks; as mass-marketed way-
finding and recreation maps; and as custom-
ized, one-off artefacts downloaded over the
Internet (Peterson, 2003) or created using
commercial, off-the-shelf mapping software
(Longley, Goodchild, Maguire and Rhind,
2005, pp. 157–75). Everyday experience sug-
gests, not without irony, that the majority of
paper maps nowadays are specially tailored
renderings based on digital data and produced
on a laser or ink-jet printer. mmSuggested reading
DeMers (2002); Longley, Goodchild, Maguire
and Rhind (2005); McMaster and Shea (1992);
Peterson (2003).digitizing The process of convertingimages,
texts,models, maps, sounds and data into
digital and computer readable forms, with an
expectation of ‘value added’ but a risk of infor-
mation loss. A paper map is easily digitized
into bitmap or JPEG format using a conven-
tional, desktop scanner. However, the user
may be disappointed by the result. First, the
original image will be converted into a series of
discrete (raster) pixels. The blocky nature of
these will be evident in words or labels copied
from the original document; the severity
will depend on the resolution of the scanner
(its dpi – dots per inch). Second, loading
the image into ageographic information
systemwill not leave it positioned correctly
with other maps and data on screen. The rea-
son is that although the digital image may
show a map, no information has been encoded
such that the computer can interpret it geo-
graphically. Providing a ‘world file’ assigns
geocodesto the corners of the image and
therefore locates it. Yet, if the user’s goal is to
select all schools within a certain distance of a
major road, for example, then s/he will still
be left wanting: the original map has been
digitized ‘en masse’ without encoding the
two specific features of schools and roads.Gregory / The Dictionary of Human Geography 9781405132879_4_D Final Proof page 163 1.4.2009 3:15pmDIGITIZING