204 Urbanization models and the regions
aColorado (USA) study found that the likelihood of development correlates bet-
terwith local patterns of existing development (described by spatial transition
models) than with the traditional factors of proximity to highways, towns, and
urban areas (Theobold and Hobbs1998).
Simulation models that mimic changing landscape patterns offer particular
promise for understanding urbanization (Franklin and Forman1987,Liet al.
1993,Swansonet al.1994,Collinge and Forman 1998 ,FormanandMellinger
2000). Modelingmosaic sequences,i.e., changing spatial patterns, permits one to
directly compare alternatives, and hence identify optimum or best options. This
approach, which will be applied to urbanization sequences later in the chapter,
is illustrated by the process for identifying the ecologically optimum spatial
sequence for changing a large landscape from a more suitable land-use type
(e.g., forest) to a less suitable type (e.g., desertified area).
First a literature review of forest cutting, suburbanization, desertification,
agricultural spread, and other broad-scale phenomena pinpointed 30 actual
mosaic sequences or changing spatial patterns in land transformation (Forman
1995,incollaboration with George F. Peterken). Five simple spatial models simu-
lated the bulk of these land transformations: (1) edge model; (2) corridor model;
(3) nucleus model; (4) few-nuclei model; and (5) dispersed-patches model. These
mosaic-sequence models were directly compared by recording levels of several
spatial attributes of ecological importance related to patch size, connectivity,
and boundary length in a hypothetical landscape.
Theedge model,whereby parallel strips are progressively degraded from one
side to the opposite side of the landscape, was found to be the best ecologically
(it retained attributes of the more-suitable initial land type furthest through
theland transformation process). In contrast, thedispersed-patches model,where
degradation occurs in small patches evenly and progressively dispersed across
theland, was worst ecologically. These results apply to most of the actual land-
use transformations observed in landscapes worldwide.
However, the edge model is not ideal. Two changes effectively create an eco-
logically better ‘‘jaws model” (Forman1995,Forman and Mellinger 2000). First,
instead of degrading parallel strips progressively from one side of the landscape,
L-shaped strips beginning from two adjacent sides are progressively degraded
towardtheopposite corner (analogous to wide-open jaws moving across the
land). Second, instead of creating a progressively larger continuous degraded
area, scattered small patches and corridors to be protected across the land are
established at the outset, and remain until the final phase.
Finally, one improvement in the jaws model makes what currently seems to be
the ecologically optimum model for shaping landscape change, the so-calledjaws-
and-chunks model(Forman and Collinge1996,Forman2002a). In the middle phase