Time-space dependencies in land-use successions at the scale of an agricultural landscape

International audienceThe agricultural landscape can be seen as an assemblage of farm territories. The way farmers organize these territories is a time AND spatial process. Understanding how a land-use succession (LUS) in a parcel depends on LUS of the neighbouring parcels is a milestone to understand the time-spatial organization of the landscape mosaic. In this work, we analyse these time-space dependencies at agricultural landscape scales. We have performed a data mining process based on hidden Markov models (HHM) to identify spatial clusters of similar distributions of LUS in 2 neighbouring parcels, furthermore called cliques. We applied this data mining process to a land-use data set covering the period from 1996 to 2007 of a 350 km2 agricultural landscape located within the Niort Plain (France). To take into account the irregular neighbour system of the parcel mosaic, we used a variable depth Hilbert-Peano scan of the area covering the landscape. Through illustrative examples of tow contrasted spatial stochastic clusters, we show that considering temporal cliques gives valuable information on the neighbour system in terms of attraction between LU