An one-dimensional process-based SVAT model (Mixfor-SVAT) was developed to describe energy, water and carbon dioxide exchanges between vegetation canopy and the atmosphere at a local scale. Simulation of the energy, water and CO2 fluxes in Mixfor-SVAT is based on aggregated description of the physical and biological processes on the leaf, tree (plant) and stand levels that allows to apply this model for prediction of atmospheric fluxes for the different vegetation types from grasslands and agricultural crops to vertically structured mono-specific and mixed forest stands represented by one or by many different tree species, as well as for description of the flux partitioning among different canopy sub-layers and different tree species. Upper boundary conditions of the model are measured or predicted meteorological parameters (air temperature and humidity, wind speed, CO2 concentration, precipitation rate and global radiation) at some height above a plant canopy within the atmospheric surface layer. For simulation of exchange processes within a multi-specific forest stand Mixfor-SVAT uses both averaged and species specific biophysical parameters of the trees describing their structure (e.g. height, crown shape, stem diameter, root depth) and biological properties (e.g. leaf stomatal conductance, photosynthesis and respiration rates, etc.). Mixfor-SVAT assumes that trees of the different species are evenly distributed over some homogeneous ground surface area and that there are no differences in biophysical properties between the same tree species. Mixfor-SVAT was applied to a natural tropical rainforest in Central Sulawesi, Indonesia. The modelled H2O and CO2 fluxes were compared with results of eddy covariance flux measurements above the forest canopy for the period from October 2003 to February 2005. All necessary biophysical parameters of vegetation and soil were obtained during several intensive field campaigns in 2004-2006. Comparisons showed a good agreement between modelled and measured H2O and CO2 fluxes especially for smoothed daily flux trends. However, a large number of spikes in measured data series caused by some instrumental errors, sensor wetting, changes in the footprint or fast changes in turbulence conditions resulted in some reduction of correlation between modeled and measured fluxes (e.g. r(2) = 0.62 for CO2 and r(2) = 0.64 for H2O fluxes under friction velocity u* > 0.3 ms(-1)). The developed Mixfor-SVAT model could be applied for solutions of the different theoretical and applied tasks, e.g. to describe the response of H2O and CO2 budgets of the different forest ecosystems to environmental and land use changes in different time scales, or to recover the long-term records of H2O and CO2 fluxes in the cases of data loss or not reliable measured fluxes. (c) 2007 Elsevier B.V. All rights reserved.