Résumé
One of the main barriers in the understanding of landscape dynamics is the high spatial variability of the surface patterns (vegetation and land cover) to which ecological processes are intimately linked. The aim of this paper is to present some newly found scaling properties for forested landscapes. Furthermore, it is advocated that patchy landscapes can sometimes be self-organized by optimizing some effective functional. A neutral landscape model has been developed in order to test this self-organization hypothesis. This model was built on the basis of a simple function, called the “Hamiltonian” analogically to physical and biological systems. The Hamiltonian is then minimized to optimize the identified landscape interactions. Fully controlled data coming from five different hundred-year runs of a process-based model appeared to be self-similar over five magnitude orders, without being explicitly simulated. The neutral model is able to reproduce the studied observations and to easily model Optimal Patchy Landscapes. The limits to this parsimonious approach that requires only one parameter (the Hamiltonian slope in loglog plot) are also discussed. The links between the effective Hamiltonian and the ecological processes still need to be investigated. Finally, such landscape Hamiltonian function appears to be a fruitful theoretical framework to describe various landscapes and potentially opens the way to a more complete dynamic landscape theory.