Abstract Drying is quite an old method to remove liquid from inside of a porous material such as wood, food, paper, ceramics, building materials, textiles, granular products, pharmaceuticals and electronic devices. The kinetics of this liquid removal depends on the material properties of its solid phase as well as on pore structure. In order to optimize the drying process for good product quality and energy considerations, theoretical modelling of the involved transport phenomena is necessary. Essentially, there are two approaches of modelling, namely the continuum and the discrete approach. The traditional approach treats the partially saturated porous medium as a fictitious continuum, and transport is described by effective parameters, which depend on saturation and pore structure. The macroscopic conservation equations for mass and enthalpy are obtained by volume averaging or homogenization, which both require a length scale separation of pore phenomena and macroscopic variation of relevant state variables such as moisture content. This requirement is not fulfilled in general, so that the second approach of discrete modelling is an alternative. In this method, the porous medium is represented by a pore network, and transport phenomena are directly investigated at the pore scale. Besides modelling of drying kinetics, such network models might also be used to calculate the effective parameters of the continuous model. The aim of this project is to study the influence of pore structure on convective drying behaviour by pore network modelling under isothermal conditions. To this purpose, a literature pore network model has been extended to describe the influence of liquid viscosity and lateral vapour transfer in the gas-side boundary layer. This model was applied to two- and three-dimensional networks of different pore size distribution and pore space topology. Four different two-dimensional network structures were investigated: the first has a mono-modal pore size distribution; the remaining three all have a bimodal pore size distribution but differ in the correlated spatial arrangement of micro and macro pores. All results are given as drying curves and phase distributions during drying. It was found that, for favourable drying with a long first drying period, a bimodal pore size distribution is essential and both micro and macro pore phase must be continuous. For efficient evaporation from wet spots on the network surface, the micro pores must additionally have a good spatial distribution. The role of boundary layer thickness, and especially lateral vapour transfer in this boundary layer, was systematically investigated and assessed by comparison with literature models. The importance of liquid viscosity in comparison with capillary forces was studied by variation of mean and standard deviation of monomodal pore size distributions. For broad distributions, capillary pumping is effective, whereas for narrow distributions, the network rather dries out with a receding front. It could also be shown that for bimodal distributions, liquid viscosity is less significant for overall drying behaviour. Concerning the influence of random generation of pore radii in the 50x50 networks, drying of all network structures was investigated by the Monte Carlo method. Additionally, the merit of periodic boundary conditions to increase effective network size was studied. A limited number of three-dimensional network drying simulations has also been carried out because they can more realistically describe certain transport phenomena in drying, like capillary pumping, than modelling in only two dimensions.