This paper investigates wood fiber and cement binder composites to propose a novel method for predicting their physical and thermal properties. These materials hold significant potential for applications in wood-wool cement boards designed for ecological and passive buildings. Three composites with varying compositions and properties were studied. Initial analyses included testing the densities, porosities, and thermal conductivities of the dry composites. The materials exhibited bulk densities ranging from 0.312 to 0.392 g/cm3, true densities from 1.986 to 2.135 g/cm3, and open and total porosities of 58.0-61.8 % and 81.6-84.3 %, respectively. Their thermal conductivities, measured over temperatures from 13 to 33°C, ranged from 0.0549 to 0.0649 W/m/K, with a linear increase observed with increasing temperature. These experimentally determined properties were employed to develop and tune the micro-scale numerical model of heat transfer in bio-based composites and the method for predicting their physical and thermal properties. The model used micro-computed tomography data to capture the complex and highly anisotropic microstructure of the composites. The main novelty of the proposed method is its ability to predict the anisotropic thermal conductivity tensor of highly heterogeneous composites, including the temperature dependence of their components. Another advantage is that it can also predict other physical properties of composites, such as density and porosity. The method demonstrated high prediction accuracy, with relative errors below ±5.4 % for all materials studied, together with high computational efficiency. This method is of significant relevance for the field, as it enables the prediction and optimization of properties in similar bio-based composites, the calculation of properties in directions not typically measured (i.e., along the width and length of boards), and the generation of input data for macro- or wall-scale models. © 2025 The Authors