Sixty-four different tree-harvesting scenarios in an initially undisturbed Dipterocarp lowland rain forest stand in Sabah(Malaysia) were simulated with the rain forest growth model FORMIND. The scenarios differ in terms of theirminimum cutting diameter, logging cycle, method and intensity. The simulation results include harvest yields and theimpact on forest structure (canopy opening and changes in species composition). Multicriteria decision analysis wasused to evaluate the scenarios and identify optimum ones by applying a stochastic extension of the PROMETHEEmethod.Almost all optimum scenarios used reduced-impact logging. High cutting limits or low logging intensities could notcompensate for the high damage caused by conventional logging techniques. Five scenarios proved to be optimum for awide range of priorities concerning different forest functions. They all use reduced-impact logging and long loggingcycles (>60 years), either with a minimum cutting limit of 50 or 60 cm stem diameter, or with medium loggingintensities.