A COMPUTATIONALLY EFFICIENT FOREST SUCCESSION MODEL - DESIGN AND INITIAL TESTS

A computationally efficient forest-stand simulation model, using a concise set of state variables, is described. The model (called FLAM: Forest LAyer Model) simulates the interaction of height-class-structured tree populations on a small patch, in contrast to gap models of the JABOWA/FORET type, which model individuals. FLAM has been derived from a generalized gap model, FORSKA, by two approximations: (1) all trees of a species in a given height class have the same stem volume, leaf area, and growth increment; and (2) the distribution of tree heights within a class uniform. Promotion from one height class to another is simulated as a binomially distributed random variable; the probability of promotion is the ratio of the predicted height growth to the depth of the height class. Height-class distributions and leaf-area-density profiles generated by FLAM were 80% and 93% similar (respectively) to profiles generated by FORSKA, but FLAM ran in 5% of the c.p.u. time. The effect of variations in spatial resolution (number of height classes) and temporal resolution (number of years per time step) were tested by comparing leaf-area-density profiles from FORSKA and FLAM. The performance of FLAM was insensitive to temporal and spatial resolution over a wide range of resolutions (1-5-year time-step, 4-20 height classes). Performance deteriorated if the temporal or spatial resolution was coarser.