Southern Forests: a Journal of Forest Science, 2017; 79(1), 65-77.Rational forest management planning requires information on the present forest state and on future development. However, forest management planning in Tanzania has often been done without any information on forest development because appropriate tools are lacking. This study presents a matrix model that combines distance-independent growth and mortality models, area-based recruitment models, and allometric models for prediction of volume and biomass. In this way forest development can be simulated according to different treatment options. A shortterm (seven years) test of the matrix model using independent data from permanent sample plots showed that the overall difference between predicted and observed basal area was small (6.5%). Long-term simulations (1 000 years) with the model showed that it was able to attain, irrespective of initial conditions, similar steady-state conditions (i.e. basal area, volume and biomass of 13 m2 ha−1, 130 m3 ha−1 and 90 t ha−1, respectively), which also correspond well to biological expectations in the ‘real’ miombo woodlands of the country. The flexibility of the model as a decision-support tool was demonstrated by simulating three harvesting options aiming at different combinations of charcoal and timber production. The model complexity is well adapted to the data quality and abundance, and it is dependent on proxies of some main drivers of the dynamic processes. The development of the matrix model is a step forward facilitating better decisions in the management of miombo woodlands. However, data ranges used for calibrating the submodels are limited in time and space, and future efforts should focus on tests and recalibrations based on extended data ranges. Presently, therefore, applications of the matrix model should be limited to the data ranges of the modelling data from the Iringa and Manyara regions.