Abstract
Accurate modeling of karst genesis is an important issue to understand the spatial and temporal evolutions of transport properties in karstified carbonate rocks. The previous modeling about karst genesis in fracture networks is mainly based on the simplified 2D conduit/fracture networks without considering wormhole formation on the scale of individual conduits or fractures. In this study, we simulated the karst genesis and wormhole formation (i.e., the emergence of preferential flow pathways within individual fractures) in carbonate joint networks using a DFN model fully-discretized in 3D and compared its difference with the 2D models. The results show that, compared to the 2D modeling, the breakthrough time (time of the abrupt rise in flow rate) in the 3D models may be much less sensitive to the network connectivity because of the flow focusing during wormhole formation on the scale of individual joints with a relatively low flow rate. Also, the dimensions of dissolving networks as well as the position and the number of developed preferential pathways in the well-connected networks may be altered when switching from 2D to 3D modeling. We further provide detailed insights into the underlying mechanism governing the dynamics of wormhole formation based on two representative networks (a critically-connected network and a well-connected network). We observed that the generated incipient karst patterns are similar between 2D and 3D models if the initial flow rate is sufficiently large (i.e., no flow focusing inside individual joints before breakthrough) or small (i.e., wormhole pathway is extremely localized, but the breakthrough time is much overestimated). If the initial flow rate falls within the transition range, the dissolution dynamics are complicated and dominated by the combined effect of flow migration over the network and flow focusing on the scale of individual joints, the magnitude of which depends on the fracture roughness. This study highlights the importance of incorporating the third dimension in DFN for accurate predictions of karst network evolution