Spatial modelling of landscape-scale vegetation dynamics, Mont Do, New Caledonia

Introduction Since human settlement of New Caledonia, the natural disturbance regime has been altered, leading to a number of changes in landscape composition and structure. In particular, the vegetation on ultramafic substrate has been greatly affected by fire and other human disturbances associated with logging and mineral exploration. It is therefore pertinent to ask how the landscape may respond to present and future disturbance regimes and how these changes might relate to the peculiarities of the serpentine environment. For example, if disturbance return intervals were to increase, or if the average size of individual disturbance events were to change, how would the landscape respond? Questions such as these are not amenable to the traditional experimental approach commonly used in ecology. Instead, it is necessary to resort to the use of simulation models (see ref. 1). Simulation modelling of landscape-level dynamics has rarely been attempted for regions characterized by ultramafic substrates, perhaps due to the overriding concern of workers to identify the more site-specific roles of unusual soil chemical conditions in determining floristic and structural patterns of vegetation. The only other landscape models considering the dynamics of serpentine landscapes that we are aware of are those of Moloney et al., Wu and Levin and Moloney and Levin, which simulate the dynamics of a serpentine grassland near Santa Cruz, California. It is of both scientific and management interest to ask how broader-scale processes might operate in serpentine landscapes, and whether interactions across scales may imply landscape-level outcomes that differ from those expected under equivalent circumstances on non-ultramafic substrates. Patch type or landscape models are the most widely used modelling framework to simulate landscape dynamics over large spatial and temporal extents. Unlike gap replacement models such as JABOWA or FORET, tree demography is not modelled explicitly in such landscape models. Instead, changes in vegetation structure are indexed to time since disturbance. These models may be described as spatially implemented patch transition simulators. Landscape models have been developed that use stochastic approaches to examine the relationship between fire regimes and landscape heterogeneity as well as fire-affected landscape changes through time, many of which are described in detail in ref. 7. Although designed to explore a variety of issues in a wide range of systems, these landscape models share several common features. These include coarser temporal resolution than mechanistic models (usually 1–10 years as compared to time-steps of minutes), ability to simulate large spatial extents with multiple fire events, and the use of stochastic algorithms. Since the temporal resolutions used in such models are much coarser than those of mechanistic models, detailed processes such as the mechanics of individual ignition events or individual tree growth and mortality can not be simulated precisely over time. Thus, in landscape models, fine-scale processes are integrated across temporal scales not by simulating them directly, but by representing them as aggregated phenomena in time and space. The model that is described and tested here falls into this broad category. As noted above, the research of Moloney et al., Wu and Levin and Moloney and Levin aside, the development and application of landscape models to serpentine landscapes has been extremely limited. The spatially-realistic (sensu ref. 8) landscape model presented here is designed to explore landscape dynamics on Mont Do, a small montane reserve in the southern part of New Caledonia. The primary aim of this paper is to explore the role of fire, and the ways that changes to the fire regime may affect landscape pattern (structure and composition). Three facets of the fire regime are considered: fire frequency (return interval), the average individual event size and the flammability of the vegetation. Serpentine Ecology South African Journal of Science 97, November/December 2001 501