The paper presents a method of calculation of thermal conductivity of bio-based building composites by applying numerical approach which based on the real morphology of sample and the solution of heat conduction equation in the micro-scale. The real microstructure of composite sample was obtained from the X-ray micro tomography (microCT). Then the three-dimensional (3D) microCT scan was processed and prepared for numerical calculations in the specially developed in-house code which allowed also for decreasing the scan resolution and for arbitrary selection of considered domain. The processed scan was used to define the computational domain of representative elementary volume (REV) as well as its microstructure and materials distribution for which the heat conduction equation was solved. By applying the volume averaging technique and by using the obtained temperature distribution in the transient state in the computational domain values of thermal conductivity of bi o-based building composite in the direction of heat flow were calculated. In the numerical model the one voxel of microCT scan corresponded to the one mesh element, therefore the generated computational meshes were quite large and it resulted in a long duration of simulations and large memory requirements. To mitigate these problems and to properly define the computational domain the influence of REV size on the obtained values of thermal conductivities of bio-based building composites was investigated. As a result, the optimal size of REV was found.