Geoscience Reference
In-Depth Information
The example of Serneels and Lambin [SER 01] is interesting in this sense. They
question the ability of a statistical and spatial model to help and understand the
forces of change of land-use in the region of the national reserve of Maasai Mara in
Kenya. This region is experiencing a strong competition between different land-
uses. It covers an area from the lands in the North of the region, called “high
potential lands”, widely open to agriculture, to the reserve covered with prairies and
subject to strict conservation rules. These two areas are separated by an area that acts
as a buffer and reserve pasture for migratory species and Maasai pastoralists. Due to
changes in land administration in this area in the 1970s, a number of major
transformations of land-use have succeeded: fragmentation of parcels and large-
scale cultivation of wheat. Between 1975 and 1995 more than 8% of the vegetation
of the surface area was lost, for the benefit of agriculture by the multiplication of the
number of small farmers-owners and the simultaneous arrival of mechanized
agriculture. The authors call up satellite images to identify the types of change
leading to the decline of the vegetation cover. Three types of change are identified:
- the conversion to a mechanized agriculture on a large-scale;
- the conversion to a subsistence agriculture when permanent colonies arrived;
- an impoverishment of the vegetation cover.
The authors have built several multiple logistic regressions to explain change
associated with each of these types of change for the two periods 1975-1985 and
1985-1995 [WRI 02]. The considered spatial entities are the pixels of 100 m. The
authors integrate a formalization of space that we call “active space”, by forming the
hypothesis of the role of spatial variables such as the distance to the capital of the
region, the accessibility to roads, to villages, the distance to water, hence integrating
the logic of Von Thunen's model [HAG 65]. These variables describing the situation
of entities relative to some spatial structuring elements are combined with the
variables describing the site: the altitude, the aptitude of the land, the type of agro-
climatic area, the type of landowner, the population density and
the population variation between the two dates. The combination of all of these
variables allows the explanatory factors of change to be approached. Depending on
the type of change considered, the explanatory powers of these factors differ:
for example, the probability of change toward large scale agriculture is
very significantly explained by the whole set of variables. This is not the case of
the probability of change toward a small owners' agriculture involving water
access and the proximity of the park, which is crucial for their pasture in times of
drought.
This approach allows the authors, on the one hand, to identify the most decisive
factors to explain the spatial distribution of change according to the type of change,
