Paysages virtuels et analyse de scénarios pour évaluer les impacts ...

- crop production objectives: the allocation of crops on the field pattern of farm fulfills production
objectives set by the farmer (Thenail and Baudry, 1994), thus farm-types may be characterized by
specific crop production objectives. This is particularly true for fodder production (grass and silage
maize), supposed to fulfill the need of the cattle; these production objectives will be assessed using
the area dedicated to each crop, assuming that the yield potentials are homogeneous over the whole
catchment;
- spatial distribution of crops: to optimize the use of the farm labor time, fields implying a lot of
operations (soil work, field input and crop management, shifting or milking of cattle) are located
closer to the farmstead (Baudry et al. 2006). Farm-types may be characterized by specific mean
distances between the crop fields and the farmstead;
- soil waterlogging: some crops are sensitive to soil waterlogging (Dogliotti et al., 2003; Houet and
Hubert-Moy, 2006).
Topography was not included as a significant driving factor because of the observed gentle slopes. We
assumed that these spatial driving factors were time-invariant over the 1993-2002 period because previous
studies showed that a combination of driving factors can be steady over several decades (Aspinall, 2004),
particularly when socio-economic factors are stable over the observed period (Pocewicz et al., 2008), which
was the case. Moreover, between 1993 and 2002, the area exchanged between farmsteads was less than 50 ha
and concerned 8 of 37 farmsteads, without significantly modifying their field pattern. The significance of the
presumed spatial factors was tested by non-parametric Wilcoxon comparison tests (p=0.95) between a set of
observed annual indicators with either a target value or another of indicators.
Crop production objectives
To test the existence of a farm-type effect on crop production objectives, we compared the crop proportions
of a farm-type (expressed as percentages of the area occupied by the farm-type) to the proportions observed
on the whole catchment area.
Spatial distribution of crop fields
To test the existence of spatial structures specific to crops in the field pattern of the farmstead, we calculated
the mean distance weighted by the field area (area-weighted distance: AWD) as follows:
AWD ft ,c , y=
F ft ,c, y
∑
field=1
F ft ,c , y
∑ area field
field=1
distance field ×area field
where AWD(ft, c, y) is the AWD (distance in m) calculated with fields belonging to farm-type ft, occupied by
crop c at year y; F(ft, c, y) is the ensemble of such fields and field is one of its elements characterized by its
distance-to-farmstead distancefield (in m) and its area areafield (in m²). This integrated indicator of the crop
spatial distribution allowed us to test two hypotheses:
III. Stochastree, un modèle de successions de cultures basé sur des arbres de décision stochastique – p. 77
(1)