Abstract. Peatlands are important natural sources of atmospheric methane (CH4) emissions. The emissions are strongly influenced by the diffusion of oxygen into the soil and of CH4 from the soil to the atmosphere. This diffusion, in turn, is controlled by the structure of macropore networks. The characterization of peat pore structure and connectivity through complex network theory approaches can give insight into how the relationship between the microscale pore space properties and CH4 emissions on a macroscopic scale is shaped. The formation of anaerobic pockets, which are local hotspots of CH4 production in unsaturated peat, can also be conceptualized through a pore network approach. In this study, we extracted interconnecting macropore networks from three-dimensional X-ray micro-computed tomography (\u00b5CT) images of peat samples and evaluated local and global connectivity metrics for the networks. We also simulated the water retention characteristics of the peat samples using a pore network modeling approach and compared the simulation results with measured water retention characteristics. The results showed large differences in peat macropore structure and pore network connectivity between vertical soil layers. The macropore space was more connected and the flow paths through the peat matrix were less tortuous near the soil surface than at deeper depths. In addition, macroporosity, structural anisotropy, and average pore throat diameter decreased with depth. Narrower and more winding air-filled diffusion channels may reduce the rate of CH4 transport as the distance from the peat layer to the soil\u2013air interface increases. Hysteresis was found to affect the evolution of the volume of connected air-filled pore space in unsaturated peat. Thus, the formation of anaerobic pockets may occur in a smaller soil volume and methanogenesis may be slower when the peat is wetting compared to drying conditions. This hysteretic behavior should be taken into account in biogeochemical models to explain the hotspots and episodic spikes of CH4 emissions. The network analysis also suggests that both local and global network connectivity metrics, such as the network average clustering coefficient and closeness centrality, might serve as proxies for assessing the efficiency of CH4 diffusion in air-filled pore networks. However, the applicability of the network metrics was restricted to the high-porosity near-surface layer. The spatial extent and global continuity of the pore network and the spatial distribution of the pores may be reflected in different network metrics in contrasting ways.