The structures of the two models are shown schematically in Fig. 1. Model (a), called the General Ecosystem Simulator (GUESS; B. Smith, I.C. Prentice, S. Sitch & M.T. Sykes, unpublished), simulates the growth of individuals on a number of replicate patches, corresponding in size approximately to the maximum area of influence of one large adult individual (usually a tree) on its neighbours. Patches are independent in terms of physical resources; that is, plants on different patches do not affect one another in the capture of light or uptake of water. However, patches are assumed to be close enough together to share a common propagule pool, establishment of new saplings of each PFT after initial colonization being directly related to the reproductive output of all individuals of that PFT the previous year (the ‘spatial mass effect’; Shmida & Ellner, 1984).

    Each woody individual belongs to one PFT (cf.taxon), with its associated parameters controlling establishment, phenology, carbon allocation, allometry, survival response to low light conditions, scaling of photosynthesis and respiration rates and the limits of the climate space the PFT can occupy. Carbon taken up through photosynthesis, and remaining following deduction of respiration and reproduction costs, is partitioned among the compartments leaf mass, fine root mass and sapwood mass, subject to certain constraints (including a constant ratio of sapwood cross-sectional area to leaf area; the pipe model of Shinozaki et al. , 1964), and to the balance between light and water limitation to photosynthesis (Haxeltine & Prentice, 1996). Each simulation year, a PFT-specific proportion of leaf mass and root mass is turned over (lost to individuals), and a fixed proportion of sapwood is converted to  eartwood: relatively more for shade-intolerant PFTs. Stem diameter, crown area and plant height are related to the sum of sapwood and heartwood mass (Huang et al. , 1992; Zeide, 1993), while bole height (the minimum height reached by the crown cylinder of each tree) is controlled by a PFT-specific minimum PAR level for photosynthesis.

    Individuals are not distinguished for grasses. A layer of grass at ground level in each patch is treated as two ‘individuals’ — one each with the C3 and C4 photosynthetic  pathways. Each grass is represented by patch totals of leaf and root carbon. Partitioning of assimilated carbon is done according to the balance between water and light limitation, as for trees. Carbon uptake through photosynthesis, plant evapotranspiration and soil water content are calculated on daily (for water balance) and monthly (photosynthesis) timesteps by a coupled photosynthesis and water module derived from the BIOME3 equilibrium biosphere model (Haxeltine & Prentice, 1996). The amount of carbon fixed by each individual each year is influenced by the quantity of photosynthetically active radiation (PAR) captured and by stomatal conductance, the latter being reduced when atmospheric evapotranspirational demand exceeds the maximum transpiration rate with fully open stomata, i.e. in the presence of water stress.

Fig. 1 Two ecosystem models differing in the representation of vegetation structure and dynamics: (a) a patch model in which individuals are distinguished and compete for light and soil water with other individuals in the same patch (‘individual-based model’); (b) a model in which individual characteristics and patch differences are averaged across a larger area for each of a number of plant functional types (PFTs) (‘area-based model’). Modules dealing with determination of environmental drivers, phenology, photosynthesis and water balance, respiration, leaf and root turnover, carbon allocation and tree allometry are common to both models.

    The fraction of incoming PAR captured by each individual across its crown area is calculated daily using the Lambert–Beer law (Monsi & Saeki, 1953), which represents an exponential reduction in available light through the canopy, based on the accumulated leaf area index (LAI, the ratio of accumulated leaf area to ground area) above a given height in each patch (Prentice & Leemans, 1990). Sun angle is not directly taken into account. PAR reaching ground level, and exceeding a minimum level for assimilation, is taken up by grasses, which are assumed to cover the entire patch area (partitioned between the C3 and C4 types according to their relative LAIs). The amount of carbon available for allocation at the end of a simulation year is reduced by maintenance and growth respiration, leaf and root turnover, and a fixed fractional allocation to reproduction for mature woody plants and all grasses.

    Model formulations of establishment and mortality are based on those employed within the ‘forest gap’ model FORSKA (Leemans & Prentice, 1989; Prentice et al. , 1993). The number of new saplings of each woody PFT and in each patch each year is drawn at random from the Poisson distribution, with an expectation influenced by a PFT-specific maximum establishment rate and by the ‘propagule pool’, i.e. the amount of carbon allocated to reproduction by all individuals of the PFT at all patches the previous year. No saplings are established in a given patch if the minimum PAR level at the forest floor is below a PFT-specific threshold, which is higher for more light-demanding species.

    Mortality of individuals is stochastic and is based on the sum of a background rate, inversely related to the PFT-specific mean non-stressed longevity, and a much higher rate, imposed only when the 5-year average mean growth efficiency (the ratio of individual net annual production to leaf area) falls below a PFT-specific threshold. The latter is higher for more light-demanding species.