Abstract In the Mediterranean Basin, landscape patterns are strongly human-modified. In recent dec- ades, because of industrialisation and rural exodus, many fields have been abandoned, generating chan- ges in the landscape pattern. In this framework, I aim to study the effect of landscape pattern on landscape dynamic processes in the Mediterranean Basin using simulation models and considering that fire may interact with landscape pattern. First I generate a gradient of five artificial random landscapes. In each landscape I include four species types growing in the Mediterranean Basin, each type with different plant traits (Quercus, Pinus, Erica and Cistus types). In each landscape scenario, each species covers 30% of the landscape but with a different spatial distribution, from the coarsest-grained (L1) to the finest-grained (L5). Then, the dynamics of each of these five landscapes were simulated for 100 years using the FATELAND simulation model. Simulations were run with six fire regime scenarios in each landscape scenario (no fire, mean fire interval of 80, 40, 20, 10 and 5 years). Landscape attributes were computed for the initial and the final landscapes. As expected, the results suggest that, as expected, some species in- crease and others decrease depending on the fire re- gime. However, the results also show that different landscape structures produce different dynamics and thus that there is a clear interaction between land- scape pattern and fire regime. For instance, coarse- grained spatial patterns generate slower dynamics than fine-grained patterns, and fire-sensitive species are maintained longer under coarse-grained patterns (i.e., fragmentation accelerates extinction of fire- sensitive species). Keywords Landscape dynamics �� Spatially explicit modelling �� Spatial pattern �� Spatial autocorrelation �� Landscape grain �� Mediterranean-type ecosystems Introduction It is widely assumed that there is a strong link be- tween ecological patterns and ecological functions and processes (Turner 1989 Levin 1992), and that there may be a two-way dynamic interaction between pattern and process. Historical land uses have pro- duced a range of landscape patterns, and the impli- cations of these patterns on ecological processes may A draft version of this paper was presented at the Special Symposium ������Global Change and Landscape Fires������ held during the IALE World Congress, in Darwin, Australia, 13���17 July 2003. J. G. Pausas (&) Forest Ecology & Restoration, CEAM Fundacion �� Centro de Estudios Ambientales del Mediterraneo, �� Charles R. Darwin 14, Parc Tecnologic, ` Paterna, Valencia ` 46980, Spain e-mail: juli@ceam.es Plant Ecol (2006) 187:249���259 DOI 10.1007/s11258-006-9138-z 123 ORIGINAL PAPER Simulating Mediterranean landscape pattern and vegetation dynamics under different fire regimes Juli G. Pausas Received: 19 July 2005 / Accepted: 8 March 2006 / Published online: 21 July 2006 �� Springer Science+Business Media B.V. 2006

persist for long periods (Peterson 2002). Similarly, fire also produces long-lasting landscape patterns (Johnson 1992) furthermore, fire may filter out (Ze- dler et al. 1983) or favour invasion (D��Antonio and Vitousek 1992 Lloret et al. 2003) of different species and thus produce changes in landscape and ecosystem processes. A clear example of strong landscape changes can be found in the Mediterranean Basin with its millenary history of intensive and extensive land use (Naveh 1990). In the Mediterranean, palae- olithic people used fire to facilitate hunting and food gathering (Oakley 1955 Perles ` 1977 Trabaud 1998 Stapert and Johanse 1999) and, since then, millennia of severe pressure resulting in burning, cutting and grazing non-arable lands, and clearing, terracing, and cultivating arable areas, have created a vast array of strongly human-modified landscapes (Farina 1998). In recent decades and especially in the northern (European) rim of the Mediterranean, with the in- dustrialisation and the rural exodus, many fields have been abandoned, increasing the cover of early-suc- cession species (many of which are very flammable), and changing the landscape pattern (Moreira et al. 2001 Pausas 2004). It is assumed that this is the main driver of the increasing number of wildfires and area burned in recent decades, although the influence of climatic changes cannot be denied (Pinol �� et al. 1998 Pausas 2004). Before the 1970s, fires were few and small areas were affected after the 70s, the number of fires and the area affected increased exponentially (Pausas and Vallejo 1999 Pausas 2004). In this framework, I ask to what extent landscape patterns may determine long-term vegetation dynamics in a fire-prone environment. There is a long history of demonstrating non-random spatial patterns in plant communities (Watt 1947 Greig-Smith 1983), and several experiments (van Andel and Dueck 1982 Thorhallsdottir �� �� 1990 Stoll and Prati 2001) indicate that the spatial arrangement of competing species can affect their performance and thus have implications on ecosystem processes. However, the consequences of the spatial pattern at landscape level have been poorly addressed (de Blois et al. 2002). Silvertown et al. (1992), using a simple spatially explicit simu- lation model for a grassland community, suggested that the process of plant dynamics may be different depending on the initial spatial pattern. To what ex- tent spatial patterns may interact with disturbance remains to be addressed. Because landscape���fire interactions are complex, dynamic and non-linear, addressing these issues at large spatial and temporal scales requires the use of simulation models (Kareiva and Wennergren 1995). Many fire-landscape studies/models consider the fuel loads and other important physical fuel characteris- tics (e.g., fuel continuity, moisture, bulk density) (Finney 1999 Lops et al. 2002 Pinol �� et al. 2005), but they do not consider the process of plant regen- eration. Thus the consequences of the different composition and structure of the landscape (in the sense of the different fire response strategies) have been poorly addressed. One of the main models used for landscape dynamics, the LANDIS model (Mlade- noff and He 1999 He and Mladenoff 1999 Sturte- vant et al. 2004), which was elaborated for boreal ecosystems, is difficult to use in Mediterranean conditions because it cannot handle short fire fre- quencies, short-lived species, seed pool dynamics and fire-stimulated recruitment, processes of great importance in the Mediterranean Basin (but see Franklin et al. 2001 and Syphard and Franklin 2004 for LANDIS applications to California Mediterranean landscapes and for future improvements of LANDIS). Spatial versions of gap models have been used for testing fire management strategies and fuel connec- tivity thresholds in coniferous forests (Miller and Urban 2000a, b). In this paper I use the FATELAND model (Pausas and Ramos 2006), a spatially explicit version of the FATE model (Functional Attributes in Terrestrial Ecosys- tems, Moore and Noble 1990) elaborated for fire- prone Australian communities and later tested for fire- prone Mediterranean ecosystems (Pausas 1999b Lloret et al. 2003). FATELAND allows different species to coexist and compete in each single cell, which is an improvement over some of the previous models test- ing spatial patterns in which each spatial cell can only be occupied by a single species (e.g., Green 1989 Silvertown et al. 1992 Lavorel and Chesson 1995 Pausas 2003). The aim is to understand how the spa- tial arrangement of different species types may interact with the fire frequency in vegetated landscape (artificial fuelbreaks such as roads and urban areas are not considered). I hypothesise that landscape with different spatial pattern would have different dynamics, and species with different regeneration strategies should have different responses to the landscape �� fire interaction. Thus, a range of artificial 250 Plant Ecol (2006) 187:249���259 123