<p>
Wildland fires create spatial patterns at multiple scales. Soil attributes, such as organic matter content, infiltration capacity or nutrient availability, and vegetation properties like canopy density and community composition may change dramatically after fire. Such patterns in soil attributes in turn affect spatiotemporal variations of fuel load and influence the propagation of future fires. The mutual reinforcement of pattern and process, a form of ecological memory, allows simple cellular automata models of fire-prone landscapes to form robust patch mosaics that resemble the patterns commonly observed in real landscapes. The fact that the patch mosaic results from the system's internal dynamics only (i.e. it is self-organized) has potentially important implications for the management of fire-prone landscape systems, which focuses increasingly on the restoration of natural processes. However, it is not always clear that simulated and measured landscape patterns coincide for the right reasons. Most spatially distributed models designed to simulate long-term dynamics of fire-prone landscapes have ecological memory build-in by simulating fire propagation as a function of fuel age or time since the last fire. Experience with real fires in forest landscapes of southern Australia shows that the impact of fuel age on fire spread is strongly diminished, or even non-existent, under severe fire weather conditions. Analyses of historical fire records and meteorological observations further suggest that self-organization through the feedback of fuel age on fire spread does not apply as a general model for pattern formation in temperate forest landscapes of Australia. Rather than dismissing self-organization as of practical relevance to the functioning of fire-prone landscape systems, these findings call for the re-thinking and elaboration of the self-organization concept plus the design of dedicated (field-based) studies that quantify key relationships between patterns and processes.</p>
