Geoscience Reference
In-Depth Information
and, on the other hand, to assess whether there has been any stability of the
explanatory factors of change over the two periods. The authors conclude on the
ability of such models to help better understand change: they remain however
inquiring on the determination of the causes and underline the difficulty of
distinguishing correlations and causes, as well as determining the meaning of their
relationships. One of the main difficulties in the study of land use changes comes
from the fact that the driving factors of change have a large geographical variability.
A second example of interest should be presented here: the modeling of the
capacity of the entities of a settlement system to persist through time [SAN 97a,
FAV 98]. The analyzed system is composed of 680 settlements. With each
settlement is associated its date of apparition during a period from the 1st to the 11th
Century. Among those, some have lasted until nowadays, while others have been
abandoned. The question is to identify the factors which “explain” this difference.
The variable “to explain” is a dummy variable: the settlement persisted (1) or not
(0). The author explores the effects of the following four factors: the site, situation,
hierarchy and accumulation effects over time. The site effect is tested from the
indicators describing the soil quality, the altitude and the topography. The situation
site effect was measured by the settlement's position in relation to the road networks
of the considered period as well as the proximity of other settlements. The hierarchy
was measured by variables describing the functional and hierarchical level of the
settlement (area, quality of materials, presence of craft, of religious building, etc.).
Finally, the accumulation includes indicators of the historical context as to whether
or not there was a previous occupation [FAV 98, FAV 12]. The model implemented
is a logistic regression. It has allowed us to show the discriminating roles on its
persistence of the hierarchical level of the settlement and of its situation at a road
intersection. Everything being equal regarding these two factors, the differences in
site or of historical context or even of proximity to a larger settlement do not
differentiate the capacity of the settlement to persist through time.
In each of these two examples, the variable to be explained (the dependant
variable) describes a change. It concerns in both cases a change in status (land-use or
existence/disappearance of a settlement). The approach would have been the same in
the context of a quantitative variation (variation of the share of a category of active
population at the level of municipalities or variation of a pollution indicator at the
level of river sections for example).
Other formalizations of variables that are to be explained using the same type of
models, make the temporal dimension directly explicit. This is the case, for example,
of a model where the variable to be explained is the opening date of a brand of
supermarket at the city level [LAU 86]. The explanatory variables called up are the
size of the city, the distance to the parent company, the political color of the city and
the existence of another supermarket of the same category. In this model, the
